feat(shim): the web APIs pages die on, and abort for fetch

[pathscale]

The web APIs the 104-site corpus found pages dying on, added to WEB_API_SHIM
in apps/chuzz/src/document_loader.rs.

Based on feat/network-apis, not master — fetch and XMLHttpRequest
live there, and the second commit here extends them. Review or land that first.

Real, with nothing invented

API	sites	why it is honest
escape / unescape	1	pure string transforms with an Annex B specification
TextEncoder / TextDecoder	2	real UTF-8 both ways: surrogate pairs joined, unpaired surrogates as U+FFFD, overlong sequences rejected
AbortController / AbortSignal	2	the whole of it is a flag and a listener list; abort, timeout and any included
String.prototype.substr	—	see below

substr is not on the corpus list and cannot be. That report counts names a
page looked up and did not find; a missing method on an existing prototype
raises TypeError: not a callable function instead, an error class counted
nowhere. It surfaced from writing unescape in terms of it and watching that
throw. It is defined rather than assigned, so it does not become enumerable on
every string in the page.

Stubs, each labelled as one in the file

• Image (4 sites) reports every image as loaded, asynchronously, without
  fetching. Most constructed Images are preloaders that need only the
  callback. Code that waits for the load and then reads pixels or natural
  dimensions gets nothing, and the zero dimensions are left honest rather than
  invented for exactly that reason. Images the document references are still
  fetched and painted by the engine; this is only the JavaScript constructor.
• ResizeObserver (1 site) never fires, unlike the IntersectionObserver
  beside it. The difference is what an invented entry would have to say:
  visibility has an answer that is right for most of a page, and a size does
  not. The only entry this could deliver carries a zero contentRect, and a
  grid that divides by that width computes zero columns and renders nothing.
• Path2D (1 site) really accumulates its path; what is missing is a canvas
  context to read it.
• ShadowRoot (1 site) is declared so instanceof is answerable and
  nothing is an instance of it, which is the truthful answer for an engine that
  builds no shadow trees.

Deliberately still absent

• getComputedStyle (3 sites). A stub answering '' for every property is
  worse than the ReferenceError it replaces. Today the script throws and stops,
  which is visible; with a lying stub it continues, measures nothing, and lays
  the page out wrongly, which reads as an engine bug. Backing it with real style
  data is engine work.
• ReadableStream (2 sites). A page reaching for it wants incremental
  delivery, and a stub can only hand over everything at once or nothing.

the_lying_stubs_are_left_out fails if either appears, so a future change that
adds them has to be one that backs them with real data.

The second commit: abort reaches the network

fetch reads init.signal — an already-aborted one rejects without touching
the network, and one that aborts later settles the promise with the signal's
reason. XMLHttpRequest.abort was an empty function and now does the same.

Half of this is honest and the other half is named as what it is not: the
request keeps running, because the host has already spawned it and there is no
cancellation channel back. What it buys is the observable half, which is the
half pages depend on. aborting_in_flight_drops_the_answer asserts exactly that
boundary — the server is contacted and does reply, and the page must not see
the reply.

Tests

The shim is a JavaScript string in a Rust file that nothing else in the build
parses, so a syntax error in it is not a compile error: it is a page that
renders as if the shim were absent, on every site. document_loader::tests
evaluates it the way a page does and reads the answers back — 7 tests over the
new globals, plus 2 in net_bridge for abort. cargo test --workspace is
green, as are fmt and clippy -D warnings.

The pre-existing embedded_ui_bundle_builds_the_solid_shell_in_boa still
passes, which matters here: WEB_API_SHIM is evaluated into the browser chrome
as well as into pages, so an addition that broke Solid would break the browser
itself.

Two more commits, and what measuring them found

DOMException, top/parent/self/frames, and scrollX/scrollY at 0 are
the tail of the corpus list past the ranked table; the omissions there
(NodeList, DocumentFragment, CharacterData, Intl, ActiveXObject,
WebAssembly, define, require) are recorded in the file with the reason and
asserted by a test.

atob/btoa is the one addition nothing asked for. Re-capturing showed a page
fall from 215 nodes to 28, which reads as a regression and is not one:
String.prototype.substr let its bundle run past the first
TypeError: not a callable function, far enough to clear the server-rendered
markup and rebuild it, and then it hit atob. Four of the twelve did the same.
A missing global is only counted once something reaches it, so fixing one defect
is what surfaces the next.

Verification

The twelve sites whose logs named a global this branch supplies, re-captured
with scripts/render-check.sh against the previous run. Site ids only; the
id-to-host map is private and the logs are not committed.

Every global this branch adds is gone from all twelve logs. What is left is
exactly the deliberate omissions (getComputedStyle, ReadableStream,
NodeList, WebAssembly), two names the pages define themselves, and two newly
reached ones now that scripts run further: Headers and File.

TypeError: not a callable function fell on 3 sites (4→0, 4→2, 6→4) and rose on
one (3→5) where the bundle now runs past its first failure into new ground.

Geometry barely moves, and that is the honest result. Nodes and boxes are
identical on 7 of 12. Three gained: +34 nodes/+13 boxes, +50/+10, and one lost 2
nodes with no box change. Two are not comparable: one page served a WAF
challenge instead of itself this run (215 nodes → 28 is that, not a render
regression), and the twelfth's counts are unchanged.

So: the error class disappears, and the pages do not visibly change much
yet. That is expected here. These shims stop scripts dying; the thing standing
between a running script and a rendered page on most of this corpus is
getComputedStyle and the DOM interface constructors, both of which need real
engine data and are deliberately not faked here.

[pathscale]

Superseded by #33, which consolidates all chuzz work into one PR.