Upgrade dependencies and improve resource handling

[tylerhutcherson]

• PR implements a few dependency upgrades (redisvl) and fixes a dependabot security warning
• Updates CI/CD flow to mirror RedisVL, removing hiredis (as this is tested in RedisVL), but adding support for Python 3.13 and Redis 8
• Improves test running modularity (supporting any optional args on a pytest command)
• Adds integration tests to the CI flow with the optional flag --run-api-tests
• Updates async resource handling where appropriate
• Removes anthropic from tests in favor of OpenAI for package simplicity and dependency management

---

So this was actually a bit tricky based on a few findings:

1. The RedisConnectionFactory deprecates get_async_redis_connection in favor of _get_aredis_connection, with the latter being an async method. It also performs module validation which is good. However, if we cutover, we'd need to implement some lazy connection strategy, but then I started to get lost so I left that out for now.
2. This package as a whole doesn't do module validation or CLIENT SETINFO stuff if the user provides their own Redis client. The RedisConnectionFactory handles that, and should probably be done here somehow?
3. The failing tests seem to be linked to the order of client initialization steps, and it no longer throwing an error?

@abrookins might ask if you have a moment early next week to take a crack at this one since you helped set up the async logic in the package and also contributed that work on redisvl. This is a start though! thanks guys!

[abrookins]

1. For get_async_redis_connection, what we're going to deprecate is calling it as a sync function. So for now, users should keep using the function but will need to await it in 1.0.0.
2. In the search index classes, we validate an existing connection if you use from_existing, so I think we should do the same in RedisVL if you supply a client in init. I can release that in a point release.
3. I'll check those failing tests! Since the connection is now lazy, if you want to test a connection error, you have to do something that uses the connection. I'll update the tests.

[bsbodden]

LGTM

[tylerhutcherson]

These three tests keep getting stuck. Curious if you all can replicate? You need an Openai key in your env. And then just run make test-all locally.

[tylerhutcherson]

These three tests keep getting stuck. Curious if you all can replicate? You need an Openai key in your env. And then just run make test-all locally.

Fixed.

[abrookins]

Looks great! Just had one question about the CI setup.