10.x: the range factories over a NULL element build a range instead of answering NULL (port of #382) #388

Problem

The same defect as #382, on release/10.x. When an element argument is NULL, the translated CreateFinite, CreateUnboundedStart and CreateUnboundedEnd still build a range. CreateFinite falls into its inverted-bounds guard and gives empty. The other two pass the NULL to PostgreSQL's range constructor, which reads a NULL bound as unbounded and gives (,). So even a positive filter selects a row that has no element: DateRange.CreateUnboundedEnd(b.Visit!.From).Contains(v) is true for a booking without a visit. EF Core answers NULL for its own value-producing members (b.Visit!.From.AddDays(1)).

Rule (c)'s predicate wrapper (#357) can't catch this, because daterange(NULL, NULL, '[)') IS NOT NULL is true. GuardedFiniteRangeConstructor and RangeConstructor are the same on both lines.

Change

Cherry-pick the #382 fix with -x, with its regression test. When an element argument can be NULL, the translation becomes CASE WHEN a IS NULL OR b IS NULL THEN NULL ELSE <today's translation> END, declared nullable. Required arguments keep today's SQL. It's SQL only: in memory the call can't be made.

The probes that found it (OuterJoinProbeTests, PR #383) are on main only, so the port's own test has to cover a value, == null, IsEmpty() and Contains(v).

Release

silent: one row under ### Changed answers in ## [10.2.0], linking this issue and its entry under docs/migration.md#changed-answers-in-1020. Fixed only on release/10.x, so close it by hand (#365).