Add SQLAlchemy session backend for conversation history management

[habema]

This PR introduces SQLAlchemySession, providing a production-grade session storage backend that can connect to any database supported by SQLAlchemy (e.g., PostgreSQL, MySQL).

This implementation is based on the discussion in Issue #1328 and follows the agreed-upon architectural pattern:

• The new session store is located in src/agents/extensions/memory/.
• SQLAlchemy is added as an optional dependency, available via the [sqlalchemy] extra.
• This is a non-breaking, purely additive change.

A test has been added.

Resolves #1328.

[habema]

@seratch

The new SQLAlchemySession feature introduces an optional dependency on sqlalchemy. This creates a conflict between two goals:

1. Keeping sqlalchemy as an optional dependency that users don't need to install if they are not using SQLAlchemySession.
2. Ensuring the SQLAlchemySession feature is tested in all our CI environments, including old_version_tests.

The old_version_tests CI job does not install optional extras by default. As a result, directly importing SQLAlchemySession in the top-level agents package causes an ImportError and breaks that CI job.

Temporary Solution

As a temporary measure to allow the feature to be merged without breaking the build, the following changes have been made:

1. No Top-Level Import: SQLAlchemySession is no longer imported in the root __init__.py of the agents package. Code that uses it must now import it directly from agents.extensions.memory.sqlalchemy_session.
2. Conditional Tests: The tests for SQLAlchemySession are now skipped in environments where sqlalchemy is not installed, using pytest.importorskip("sqlalchemy").

Next Steps

This solution is a trade-off. It prevents the CI from breaking, but it means that the SQLAlchemySession feature is not currently tested in the old_version_tests environment.

I think the correct long-term fix is to update make old_version_tests to install the [sqlalchemy] extra (and any other extra dependencies needed) for all test jobs. This will allow the tests to run everywhere and provide full test coverage for the new feature.

Let me know what you think.

[seratch]

At a glance, this looks good to me; @rm-openai any thoughts?

[seratch]

@habema could you resolve the conflict?

[habema]

@seratch you got it.
On a sidenote, all mypy and linter errors I fixed are unrelated to any files I changed in this PR. Might need some looking into their behavior.

[seratch]

Overall, this enhancement looks good to me; great work 👍