🤖 I have created a release \*beep\* \*boop\* 
---
### Updating meta-information for bleeding-edge SNAPSHOT release.
---


This PR was generated with [Release Please](https://github.com/googleapis/release-please).

This PR was generated using Autosynth. 🌈


<details><summary>Log from Synthtool</summary>

```
2020-11-17 09:41:32,582 synthtool [DEBUG] > Executing /root/.cache/synthtool/java-spanner/.github/readme/synth.py.
On branch autosynth-readme
nothing to commit, working directory clean
2020-11-17 09:41:33,601 synthtool [DEBUG] > Wrote metadata to .github/readme/synth.metadata/synth.metadata.

```
</details>

Full log will be available here:
https://source.cloud.google.com/results/invocations/005ecbdd-e849-49dc-b49e-0b895e2d2872/targets

- [ ] To automatically regenerate this PR, check this box.

…643)

If the first statement of a read/write transaction was a query or a read operation,
and the application would not call ResultSet#next() on the return result, the transaction
would hang indefinetely as the query would be marked as the one that should initiate the
transaction (inline the BeginTransaction option). The query would however never be
executed, as the actual query execution is deferred until the first call to ResultSet#next().

Fixes #641

Prevents the re-generation of the github CODEOWNERS file.

…to v3.0.4 (#636)

…2.0.8 (#644)

This PR was generated using Autosynth. 🌈


<details><summary>Log from Synthtool</summary>

```
2020-11-18 00:20:16,991 synthtool [DEBUG] > Executing /root/.cache/synthtool/java-spanner/.github/readme/synth.py.
On branch autosynth-readme
nothing to commit, working directory clean
2020-11-18 00:20:17,985 synthtool [DEBUG] > Wrote metadata to .github/readme/synth.metadata/synth.metadata.

```
</details>

Full log will be available here:
https://source.cloud.google.com/results/invocations/44f2d200-ccfc-470d-8d23-b1d174cc7c25/targets

- [ ] To automatically regenerate this PR, check this box.

The query_ThenUpdate_ThenConsumeResultSet did not take into account that the session pool
initialization will execute 4 BatchCreateSessions requests. These will normally be executed
before the test transaction, but sometimes at least one of those requests might arrive after
the transaction has started. That could cause the last 3 requests on the mock server to be
different from what the test expected.

Co-authored-by: release-please[bot] <55107282+release-please[bot]@users.noreply.github.com>