fix(extensions): set-priority repairs corrupted boolean priority

[jawwad-ali]

Description

The extension set-priority command has a skip-if-unchanged guard whose comment promises that corrupted values get repaired:

# This ensures corrupted values (e.g., "high") get repaired even when setting to default (10)
if isinstance(raw_priority, int) and raw_priority == priority:
    console.print(f"...already has priority {priority}")
    raise typer.Exit(0)

But in Python isinstance(True, int) is True and True == 1 (False == 0). So a corrupted boolean priority (e.g. True) matches the guard when the user runs set-priority <ext> 1: the command reports "already has priority 1" and exits without rewriting the bool to a real int — the exact opposite of what the comment promises. normalize_priority() already guards this (if isinstance(value, bool): return default); this call site didn't.

Fix

Exclude bools from the skip guard, mirroring normalize_priority:

if (
    isinstance(raw_priority, int)
    and not isinstance(raw_priority, bool)
    and raw_priority == priority
):

Now a corrupted bool falls through and is repaired to a real int.

Testing

• New test_set_priority_repairs_corrupted_bool: installs an extension, injects priority: True, runs set-priority test-ext 1, asserts it reports priority changed (not already has priority) and the reloaded value is 1 with not isinstance(..., bool). Fails before (reports "already has priority 1", verified by source-stash), passes after.
• TestExtensionPriorityCLI passes (8 tests); uvx ruff check clean. The existing same-value test (priority 5, a real int) stays green.

AI Disclosure

• I did use AI assistance (describe below)

Found and fixed with Claude Code (Claude Opus 4.8) under my direction. AI flagged the bool-is-int trap against the comment's promise; I confirmed the short-circuit, proved fail-before via source-stash, and reviewed the diff before submitting.

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.