test: add standalone e2e test cases for deterministic recovery test

[yezizp2012]

Is your feature request related to a problem? Please describe.

Currently we will run deterministic recovery test as follows for all PRs:

echo "--- deterministic simulation e2e, ci-3cn-1fe, recovery, streaming"
seq 1 | parallel MADSIM_TEST_SEED={} $RUNNER --kill-meta --kill-frontend --kill-compute './e2e_test/streaming/\*\*/\*.slt'

# bugs here! Tracking issue https://github.com/risingwavelabs/risingwave/issues/4527
echo "--- deterministic simulation e2e, ci-3cn-1fe, recovery, batch"
seq 1 | parallel MADSIM_TEST_SEED={} $RUNNER --kill-meta --kill-frontend --kill-compute './e2e_test/batch/\*\*/\*.slt'

But:

1. there're so many test cases under ./e2e_test/streaming/, it will definitely timeout since the test will randomly kill components during running each sql except DML.
2. the test is soft pass now because some bugs are still under fixing. This can lead to new bugs not being found when they appear, which has happened more than once.

Describe the solution you'd like

We should sort out some e2e test cases that recovery tests may involve, the cases should includes:

1. ddl operations, including table/stream sources.
2. cases that cover almost all Executors to test whether they will recovery properly.
3. data consistency, including check of the ability to handle dml correctly after recovery.

After that, we should force enable the deterministic recovery test to check whether this PR will break any recovery process.

Describe alternatives you've considered

No response

Additional context

No response

[yezizp2012]

Cc @wangrunji0408

[yezizp2012]

Currently only dynamic_filter.slt, outer_join.slt and tpch_upstream.slt will fail the deterministic recovery test, the bugs are recorded in #5726 #6093 .