🤖 I have created a release *beep* *boop*
---


### Updating meta-information for bleeding-edge SNAPSHOT release.

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

…to v6.65.1 (#3070)

* chore(deps): update dependency com.google.cloud:google-cloud-spanner to v6.65.1

* 🦉 Updates from OwlBot post-processor

See https://github.com/googleapis/repo-automation-bots/blob/main/packages/owl-bot/README.md

---------

Co-authored-by: Owl Bot <gcf-owl-bot[bot]@users.noreply.github.com>

* chore: propagate otel context in Connection API

Use a TaskWrapping executor in the Connection API to propagate the
current OpenTelemetry context to the executor thread. This ensures
that the context that is used in the end-user application is
propagated to the Spanner client library, and tracing shows a full
view of the spans from the application to the gRPC layer.

* 🦉 Updates from OwlBot post-processor

See https://github.com/googleapis/repo-automation-bots/blob/main/packages/owl-bot/README.md

---------

Co-authored-by: Owl Bot <gcf-owl-bot[bot]@users.noreply.github.com>

* chore: try with random channel hint

* chore: add option for random channel

* chore: actually use random channel option

* chore: only lock the specific wrapper

* chore: simplify creation and assignment

* chore: make more variables final

* chore: use separate pool

* chore: use a separate mux client

* chore: make init blocking

* chore: disable pending tx check

* chore: add call durations to client lib

* chore: add call_durations

* chore: use session pool for mux session

* chore: use mux database client

* chore: make mux client optional

* refactor: move multiplexed session handling to separate class

* chore: cleanup

* feat: add maintainer

* chore: add more tests

* chore: fix test failures

* fix: ChannelUsageTest should keep session in use for longer

* test: skip ChannelUsageTest in all cases

* chore: keep track of DatabaseDeleted errors

* fix: freeze server to prevent flakiness

* fix: freeze server to prevent flakiness

* test: add retry loop for second query

* chore: remove print statements and add emulator handling in test

* test: add tests for maintainer

* chore: register SessionConsumer for reflection

* chore: skip test as it fails on native build

Just skip the entire test, as the scenario is already covered by
other tests, and spending time on trying to figure out what is
happening in this failure is just a waste of time.

* 🦉 Updates from OwlBot post-processor

See https://github.com/googleapis/repo-automation-bots/blob/main/packages/owl-bot/README.md

* chore: add random channel hint as option

* chore: add single-use channel hint

* 🦉 Updates from OwlBot post-processor

See https://github.com/googleapis/repo-automation-bots/blob/main/packages/owl-bot/README.md

* chore: single-use hint

* 🦉 Updates from OwlBot post-processor

See https://github.com/googleapis/repo-automation-bots/blob/main/packages/owl-bot/README.md

* chore: use next available channel

* chore: keep track of num transactions and channels in use

* chore: remove println

* chore: remove option for using session pool for mux

* feat: add UNIMPLEMENTED handler

* chore: cleanup

* chore: add TODO for removing the unimplemented handling

* 🦉 Updates from OwlBot post-processor

See https://github.com/googleapis/repo-automation-bots/blob/main/packages/owl-bot/README.md

* chore: run formatter

* test: fix flaky tests

Fixes #3050
Fixes #3081
Fixes #3080

---------

Co-authored-by: Owl Bot <gcf-owl-bot[bot]@users.noreply.github.com>

Adds an environment variable that can be used to force-disable the
use of multiplexed sessions.

Adds support for running DDL statements when a connection is in autocommit=false mode. By default, DDL statements are only allowed when no transaction is active. That is; no query or DML statement has been executed which activated a read/write transaction.

A new flag is added that can be used to revert the behavior back to the original behavior where DDL is always refused when autocommit=false. The same flag can also be used to make the API behave the same as MySQL and Oracle, where any active transaction is automatically committed whenever a DDL statement is encountered.

Concretely this means that the following is now allowed:

```
set autocommit=false;
create table Singers (SingerId INT64, Name STRING(MAX)) PRIMARY KEY (SingerId);
```

The following is by default NOT allowed, unless
ddlInTransactionMode=AUTO_COMMIT_TRANSACTION

```
set autocommit=false;
select * from singers; -- This starts a transaction
create table Albums (AlbumId INT64) PRIMARY KEY (AlbumId); -- This is not allowed
```

Fixes #3084

Uses the gson Java version checker, which is more production-hardened than the custom one in the Spanner client. It seems that the Spanner implementation does not get it right on all versions on Windows.

…3088)

The metrics for multiplexed sessions were not included in the
refactoring that moved multiplexed sessions out of the session pool.
This change re-adds those metrics based on the new client for
multiplexed sessions.

The CreateSession calls are irrelevant to the test, as multiplexed sessions are not used for read/write transactions.

Fixes #3090

Include the stack trace of where a session was checked out when the session pool is exhausted. This makes it easier to find the place in application code that is responsible for having checked out sessions, and potentially having leaked these.

…v3.30.0 (#3082)

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