add-variants: fix .tbi/.csi crashing the enricher (v2.8.2) + MGBOA license

[lucapinello]

Summary

Two things, bundled because the version bump is shared:

1. Bug fix — add-variants crashes on tabix/bcftools index sidecars

The VCF-directory listing filtered .tbi with a mutate-while-iterating loop:

listChrs = os.listdir(dirVCFFiles)
for file in listChrs:
    if file.endswith('.tbi'): listChrs.remove(file)   # skips elements!

Removing from a list during iteration skips elements, so some .tbi files survive and are passed to the enricher, which gzip.open()s + readline()s each input → decompressing a binary .tbi as text raises UnicodeDecodeError: 'utf-8' codec can't decode byte 0x9d. Replaced with a single comprehension excluding .tbi and .csi:

listChrs = [f for f in os.listdir(dirVCFFiles) if not (f.endswith('.tbi') or f.endswith('.csi'))]

Surfaced building a genotyped 1000G index whose VCF folder was tabix-indexed. The stray index workers were harmless (they died before writing) but produced confusing tracebacks and were a latent collision risk.

Verified (independent review):

• Reproduced the survivor + the enricher crash.
• The other listChrs (genome dir, indexGenome) uses continue, not remove — no twin bug.
• Nothing downstream needs the index entries (len(listChrs) worker count + the for elem in listChrs enrichment loop want real VCFs only).
• The enricher hard-requires gzip input, so dropping index sidecars is necessary and non-regressing (a .csi/.tbi reaching it would crash the same way).

2. Version + license

• VERSION 2.8.1 → 2.8.2 (crispritz.py + conda/meta.yaml) + CHANGELOG entry.
• LICENSE: AGPL-3.0 (dual) → MGB Open Access License 1.0 (MGBOA 1.0), matching CRISPRme+ and CRISPRme (non-commercial academic use; commercial use requires a commercial license).

Note: the stale conda/crispritz.py and docker_image/crispritz.py copies (not installed by conda/build.sh or the root Dockerfile, which both use root crispritz.py) are left untouched.

🤖 Generated with Claude Code