Distinguish unnamed timezone file contents in equality

[FanWu-ai]

Pull Request Check List

• Added tests for changed code.
• Updated documentation for changed code.

Fix

Unnamed TZif-backed timezones all have key=None, so the current key-only equality makes distinct Paris/UTC/New York files equal and collapses them into one set/dictionary entry. This also affects the existing local-timezone fallback for copied /etc/localtime files.

Compare and hash the TZif bytes already retained for pickling when the key is absent. Independent loads of identical bytes and pickle round trips remain equal. Named/keyed timezone behavior and FixedTimezone are unchanged.

Twelve new cases cover distinct and identical files, collections, all pickle protocols, named/keyed/uncached controls, unrelated objects and copied-localtime discovery. Documentation explains the rule.

Validation

Linux/Python 3.12.14, unchanged dependencies and pinned hooks:

• Fresh complete baseline: 1,871 passed / 3 existing skips in each pure-Python/native mode.
• Final source and installed-wheel complete suites: 1,883 passed / 3 existing skips in each mode.
• Exact regressions on pristine source: 4 failures / 13 controls per backend.
• Strict mypy (228 files), all six hooks including Rust formatting/Clippy, strict MkDocs, rebuilt release wheel/sdist and strict Twine pass.
• Artifact production-source bytes and native import provenance verified. Independent full reruns and 3,865 additional checks per backend pass across 23 zones, aliases, collection behavior, pickle protocols and copies.

Existing Windows/timezone skips, INFO-level docs anchor notices and Cargo metadata warning are unchanged. Other Python/OS combinations and hosted CI are not claimed passed. Prepared and tested with AI assistance (OpenAI Codex); no human review is claimed.

[codspeed]

Merging this PR will not alter performance

⚠️ 1 benchmark spent significant time in system calls

System calls cannot be consistently instrumented, so they are not included in the measure, which understates the real cost. Please switch to the Walltime instrument to accurately measure system calls.

Measurement and system calls

• test_parse_iso8601

ℹ️ 1 benchmark has no flame graph

The flamegraph could not be generated because we failed to resolve the symbols in the profiling trace. This happens when we fail to harvest information about language frames.

• test_parse_iso8601

✅ 1 untouched benchmark

---

_{Comparing FanWu-ai:fix-unnamed-timezone-equality-20261012 (beb8f70) with master (7840d47)}

[Wkhanex]

The content-based distinction also holds when two unnamed zones happen to share the current UTC offset. On beb8f70, an unnamed Europe/London zone and an unnamed UTC zone both have offset zero in January 2024, but remain unequal and occupy two set entries, preserving their different rule histories. I additionally checked UTC, Paris, New York and Kathmandu through every available pickle protocol: 24 round trips preserved equality, hash and dictionary lookup, and January/July offsets matched ZoneInfo. The timezone module passed 46 tests with its two existing POSIX-transition skips. Python 3.12/pure-Python mode only; no full-suite or native-backend claim.

Review and local validation prepared with OpenAI Codex assistance.