Fix %aunique{} not disambiguating sanitized album name collisions

[marubio04]

Fixes #6462

Summary

This fixes %aunique{} disambiguation for albums whose raw names differ but collide after path sanitization, such as 1/1 and 1?1.

Changes

• Detect album path collisions using sanitized path-formatted key values.
• Add a regression test for sanitized album-name collisions.
• Add a changelog entry.

Testing

• poe test
• poe lint
• poe check-format

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 71.79%. Comparing base (8a82a92) to head (327b985).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #6582      +/-   ##
==========================================
+ Coverage   71.78%   71.79%   +0.01%     
==========================================
  Files         159      159              
  Lines       20526    20535       +9     
  Branches     3264     3266       +2     
==========================================
+ Hits        14734    14743       +9     
  Misses       5104     5104              
  Partials      688      688              

Files with missing lines	Coverage Δ
beets/library/models.py	87.19% <100.00%> (+0.17%)	⬆️

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[Copilot]

Pull request overview

PR fix %aunique{} so it disambiguate albums that different in raw metadata but become same after path sanitization (ex: 1/1 vs 1?1). This fit beets path formatting + import/move flow where album dirs must not collide.

Changes:

• Update %aunique{} collision set to consider path-sanitized values, and sanitize disambiguator values when picking a field.
• Add regression test for sanitized album-name collision.
• Add changelog entry for bug 6462.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 5 comments.

File	Description
beets/library/models.py	Teach unique template logic to consider sanitized collisions and sanitized disambiguator values.
test/test_library.py	Add test covering sanitized album name collision case.
docs/changelog.rst	Document bugfix (and adds an extra 6450 bullet too).

[Copilot]

grug worry sanitize_path here also ignore library custom replacements. if user set replace rules, disambiguator uniqueness check might pick field that look unique pre-replace but collide on disk.

grug suggest pass replacements=self.lib.replacements into sanitize_path.

[Copilot]

grug see perf danger. this code do self.lib.albums() full scan inside %aunique{} compute. big library => many work, maybe O(N^2) over many albums.

[Copilot]

grug also note formatted(for_path=True) default include_keys='*' trigger extra db query for keys (see FormattedMapping). inside full scan, that hurt a lot. grug suggest use item.formatted(included_keys=keys, for_path=True) (and maybe skip scan when sanitize no change).

Suggested change

	formatted = item.formatted(for_path=True)
	formatted = item.formatted(
	included_keys=keys,
	for_path=True,
	)

[Copilot]

grug see changelog add 🐛6450 entry here, but this PR diff show no code for that bug. also same 6450 line already exist later in file, so this look like duplicate / wrong backport.

grug suggest drop this 6450 bullet (or move it only if PR actually include that fix).

[Copilot]

grug worry sanitize_path here ignore library custom replacements. real path build use replacements=self._db.replacements, so collision detect may disagree with actual path when user set replace rules.

grug suggest pass replacements=self.lib.replacements (or self._db.replacements) into sanitize_path call.

Suggested change

	util.sanitize_path(str(formatted.get(key, "")))
	util.sanitize_path(
	str(formatted.get(key, "")),
	replacements=self.lib.replacements,
	)

[ShimmerGlass]

Hey @marubio04!

Could you please give details on the performance impact of this change?

For mid to large libraries listing and computing paths for all albums can take seconds, or tens of seconds. If I'm not mistaken with this change this will happen each time we want the path of an item (excluding cases when conflicts are dectected with the query, which should be very rare). This could dramatically slow beets down for day to day operations.