fix(prefilter): use token-boundary matching for test and migration framework detection

[Adityakk9031]

Summary

• Replaces unbounded substring matching in has_test_patterns and has_migration_patterns with token-boundary aware matching matches_token_or_module.
• Prevents false-positive test file exclusions on legitimate production files that import modules containing substrings of test words (such as \�ttestation, \contest, \latest, \astest, \user_migration).
• Adds regression unit test cases in \ ests/unit/prefilter_should_scan_tests.rs.

[juangaitanv]

vitest, Go testing, and testify no longer match the test patterns, so their imports are scanned. could we retain explicit patterns for these framework roots while preserving the boundary fix?

[Adityakk9031]

@juangaitanv Thanks for the review! Added explicit patterns for "testing", "testify", "vitest", and "github.com/stretchr/testify" to test_patterns while retaining token boundary checking. Added unit test file_with_testing_or_vitest_import_is_skipped.

[juangaitanv]

from django.db import migrations no longer matches a migration pattern, so those files are scanned in general mode. could we retain generic migration indicators while treating _ and digits as identifier characters?

[Adityakk9031]

@juangaitanv Restored "migration" and "migrations" to migration_patterns, and updated is_identifier_character to treat [A-Za-z0-9_] as non-boundary identifier characters. Now from django.db import migrations matches and is skipped, while identifiers like attestation or user_migration are not falsely matched. Added unit tests for both cases.