ci: split Tests workflow into 4 parallel shards via pytest-split

Reduces CI wall time by running the test suite as 4 parallel matrix
jobs instead of a single job. Each shard runs ~3,000 tests in
parallel, so total wall time drops from ~4min to ~60-90s.

Changes:
- Add pytest-split to dev extras (deterministic test splitting,
  composes with pytest-xdist's -n auto inside each shard).
- Matrix-split tests.yml 'test' job into 4 groups. Each shard runs
  'pytest ... --splits 4 --group N' and parallelizes inside with
  the -n auto already in pyproject.toml's addopts.
- fail-fast: false so all shards finish even if one fails
  (consistent with current behavior when there's no matrix).

Expected CI timing:
  Before: 243s single-job (4m03s)
  After:  ~60-90s per shard in parallel + ~25s install overhead
          \u2192 total CI ~90-115s

No test-file changes. Deterministic hash-based distribution (no
.test_durations file yet; can add one later for better balance).

The e2e job is unchanged — it's already small (20s) and runs
separately.

.github/workflows/tests.yml

@@ -16,8 +16,13 @@ concurrency:
 
 jobs:
   test:
+    name: test (${{ matrix.group }}/4)
     runs-on: ubuntu-latest
     timeout-minutes: 10
+    strategy:
+      fail-fast: false
+      matrix:
+        group: [1, 2, 3, 4]
     steps:
       - name: Checkout code
         uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5  # v4
@@ -37,10 +42,11 @@ jobs:
           source .venv/bin/activate
           uv pip install -e ".[all,dev]"
 
-      - name: Run tests
+      - name: Run tests (shard ${{ matrix.group }}/4)
         run: |
           source .venv/bin/activate
-          python -m pytest tests/ -q --ignore=tests/integration --ignore=tests/e2e --tb=short -n auto
+          python -m pytest tests/ -q --ignore=tests/integration --ignore=tests/e2e --tb=short \
+            --splits 4 --group ${{ matrix.group }}
         env:
           # Ensure tests don't accidentally call real APIs
           OPENROUTER_API_KEY: ""