audit_predicate_domains crashes on a non-UTF-8 locale (report/baseline I/O)

[realmarcin]

Found in the self-review of PR #314 and fixed in that same PR (commit pins
encoding="utf-8"); filed as the record per the review-finding policy.

scripts/audit_predicate_domains.py writes a non-ASCII ⊑ (U+2291, "⊑ microbe")
into the detail column of 366 report rows. The report write, baseline write,
and baseline read all used the platform default (locale) encoding. On a
non-UTF-8 locale that raises UnicodeEncodeError writing the report and
UnicodeDecodeError reading the baseline — a hard crash. It only passed CI
because the GitHub runner's locale is UTF-8.

Fix: encoding="utf-8" on all three opens. Verified byte-neutral — the
committed report and baseline are byte-identical after regeneration.

Related (NOT fixed here, pre-existing repo-wide): the corpus read
yaml.safe_load(path.read_text()) at line ~143 also relies on the default
encoding, matching the sibling audits (audit_causal_graphs.py etc.). If a
trait YAML ever carries non-ASCII (e.g. an accented species name) it has the
same latent failure across every audit. Worth a separate, repo-wide sweep rather
than a one-file change here.

[realmarcin]

Fixed in PR #314 (encoding pinned to UTF-8; verified byte-neutral). Closing as the audit-trail record; the related pre-existing corpus-read pattern is left noted for a separate repo-wide sweep.