chore(bigtable): fix flaky sharded query concurrency limit test by relaxing eps (#17541)

**Flaky tests**:

- The test `test_read_rows_sharded_concurrency_limit` was flaky under VM
execution load because it checked if all of the first 10 concurrent
requests were dispatched within `eps = 0.01` seconds (10 milliseconds)
of the operation's start.
Due to CPU scheduling and thread context switching overhead on
virtualized CI hosts (like Kokoro), dispatching 10 concurrent
threads/tasks can occasionally take slightly longer than 10ms,
triggering random `AssertionError` failures.

**Solution**:

- Relaxed the threshold `eps` from `0.01` to `0.2` seconds (200
milliseconds) in:
- The async source test file:
[tests/unit/data/_async/test_client.py](file:///usr/local/google/home/omairn/git/googleapis/google-cloud-python/packages/google-cloud-bigtable/tests/unit/data/_async/test_client.py)
- The sync auto-generated test file:
[tests/unit/data/_sync_autogen/test_client.py](file:///usr/local/google/home/omairn/git/googleapis/google-cloud-python/packages/google-cloud-bigtable/tests/unit/data/_sync_autogen/test_client.py)
This allows enough scheduling margin for virtualized CI runners to pass
successfully while still validating that the first 10 queries are fired
concurrently without delay compared to the queued queries.

Author: ohmayr

packages/google-cloud-bigtable/tests/unit/data/_async/test_client.py

@@ -2328,7 +2328,7 @@ async def mock_call(*args, **kwargs):
                         starting_timeout - kwargs["operation_timeout"]
                         for _, kwargs in read_rows.call_args_list
                     ]
-                    eps = 0.01
+                    eps = 0.2
                     # first 10 should start immediately
                     assert all(
                         rpc_start_list[i] < eps for i in range(_CONCURRENCY_LIMIT)

packages/google-cloud-bigtable/tests/unit/data/_sync_autogen/test_client.py

@@ -1926,7 +1926,7 @@ def mock_call(*args, **kwargs):
                         starting_timeout - kwargs["operation_timeout"]
                         for _, kwargs in read_rows.call_args_list
                     ]
-                    eps = 0.01
+                    eps = 0.2
                     assert all(
                         (rpc_start_list[i] < eps for i in range(_CONCURRENCY_LIMIT))
                     )