[Cosmos] add a true end to end timeout

[simorenoh]

The Python SDK currently exposes several different timeout and connection retry configurations that users can leverage: https://github.com/Azure/azure-sdk-for-python/blob/main/sdk/cosmos/azure-cosmos/docs/TimeoutAndRetriesConfig.md

However, all of these configs only deal with network latency and the end to end of the request over the wire. That is, the timers keeping track only care about the request being sent and the response being returned. Since for some operations, like vector search queries, the client can take additional time in formatting the results before returning them to the user, the end-to-end request from the user's experience might go over the timeout they're expecting.

The ask here is to introduce a true end-to-end timeout, that takes into account the additional time a request might spend within the client's flow or multiple requests on the server side and throw an exception if that limit is reached.

[annatisch]

Rather than introducing yet another timeout - the client-side code should be amended to respect the existing timeout param? Otherwise it would difficult to construct customer code that cleanly combines the two together.

The assumption when the timeout feature was added to the core pipeline was that the client code that surrounds the pipeline is so minimal that it could be generally disregarded in a timeout.
The Cosmos SDK would be an exception to that assumption, so we should take the extra plumbing steps required so that the same contract still applies.