e2e tests: limit name_len_slow to 3, split e2e tests from other tests

[The `test_query_e2e` takes almost ~8mins to run][1] (whole CI job takes
11 mins). The `name_len_slow` script is the main culprit, since it
sleeps for 1 sec in each udf function and that mapper is run in a single
process parallel mode.

```
474.21s call     tests/test_query_e2e.py::test_query_e2e@tmpfile
```

This commit adds a limit of 3 files to the name_len_slow script, which is
enough, since it's only running a single process.
(We immediately interrupt the running process after seeing "UDF
Processing Started" gets printed).

This also split tests into two: one for the e2e tests and one for the
rest, so that these things are more obvious in the future.

[1]: https://github.com/iterative/datachain/actions/runs/12879531971/job/35907168617#step:8:82

.github/workflows/tests.yml

@@ -112,7 +112,11 @@ jobs:
         run: echo 'DISABLE_REMOTES_ARG=--disable-remotes=azure,gs' >> $env:GITHUB_ENV
 
       - name: Run tests
-        run: nox -s tests-${{ matrix.pyv }} -- $DISABLE_REMOTES_ARG
+        run: nox -s tests-${{ matrix.pyv }} -- -m "not e2e" $DISABLE_REMOTES_ARG
+        shell: bash
+
+      - name: Run E2E tests
+        run: nox -s tests-${{ matrix.pyv }} -- -m "e2e" $DISABLE_REMOTES_ARG
         shell: bash
 
       - name: Upload coverage report

tests/scripts/name_len_slow.py

@@ -35,6 +35,6 @@ def name_len(file):
 DataChain.from_storage(
     "gs://dvcx-datalakes/dogs-and-cats/",
     anon=True,
-).filter(C("file.path").glob("*cat*")).settings(parallel=1).map(
+).filter(C("file.path").glob("*cat*")).limit(3).settings(parallel=1).map(
     name_len, params=["file"], output={"name_len": int}
 ).save("name_len")