test: drop the cwd-relative sys.path.insert calls from the test suite

[yuneng-berri]

TLDR

Problem this solves:

• 1,015 test files insert a path that never points at litellm
• The argument resolves against the working directory, not the file
• From the repo root it names the directory above the checkout
• TQ003 has sat at its 1,077 seed since the rule shipped

How it solves it:

• Delete those calls, and the imports they were the only user of
• Ratchet the TQ003 ceiling from 1,077 to 62

User Flow

No end-user behavior changes and no test behavior changes. A contributor copying
an existing test file as a template stops inheriting a line that does nothing.

Relevant issues

Linear ticket

Pre-Submission checklist

• I have added meaningful tests
• The handful of test files covering my change pass locally
• My PR passes all required CI/CD checks (e.g., lint, schema.d.ts sync check, etc.)
• My PR's scope is as isolated as possible; it only solves 1 specific problem
• I have received a Greptile Confidence Score of at least 4/5 before requesting a maintainer review

Screenshots / Proof of Fix

Before (ff02d5c)

What the line actually inserts

1. Every job runs pytest from the repo root, so:

python -c "import os; print(os.path.abspath('../..'))"
/Users/yunengjiang/development

2. That is the directory above the checkout, not litellm. The import works
   because the package is installed into the environment

The rule's count and shape

1. python scripts/check_test_quality.py tests | grep -c ' TQ003 '

1077

2. Grouped by the argument, 1,015 of the 1,077 resolve against the working
   directory:

  460  os.path.abspath('../..')
  211  os.path.abspath('../../..')
  153  os.path.abspath('../../../../..')
  128  os.path.abspath('../../../..')
   24  os.path.abspath('../../../../../..')
   15  os.path.abspath('../../../')
    ...

Collection baseline

1. pytest <every touched tree> --collect-only -q

45871 tests collected, 51 errors in 8.91s

After (a704dfb)

The count drops to the calls that resolve against file

1. python scripts/check_test_quality.py tests | grep -c ' TQ003 '

62

2. Those 62 use os.path.dirname(__file__), Path(__file__) or a variable, so
   they are a different question and are untouched

Collection is unchanged

1. Same command on this branch:

45871 tests collected, 51 errors in 23.02s

2. diff of the two runs' ERROR lines is empty, so the same 51 pre-existing
   collection errors and no new ones

No new undefined names

1. ruff check --select F821 tests/ before and after
2. Both report the same single pre-existing finding, so nothing lost an import
   it was still using. One file keeps its import os because another test module
   imports the name back out of it

The gates agree

OK: every TQ rule is within its test-suite ceiling (base origin/litellm_internal_staging)
OK: every strict rule is within its codebase ceiling (base origin/litellm_internal_staging)
OK: every LIT rule is within its codebase ceiling (base origin/litellm_internal_staging)

Type

🧹 Refactoring
✅ Test

Caveats (if any)

• 998 files, so it will conflict with anything in flight
• Every removal is one of two shapes, so the diff reads at a glance
• 62 __file__-relative call sites are deliberately left
• Greptile skipped it: 999 files, past its 100 file limit

Final Attestation

• The tests check the right things, including the edge cases, and regressions in the respective real-world customer use-cases are not possible after this PR

[greptile-apps]

Too many files changed for review (999 files, 100 file limit).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!