Add shard awareness

[sylwiaszunejko]

Added support for shard awareness:

• open connections to every shard from every host
• use shard aware info when routing queries.

Fixes: #10 #3

[dkropachev]

Can you please create a test to check if pool is getting filled with connections when connections are disconnecte or become broken.

[sylwiaszunejko]

@dkropachev the last commit makes the driver update the sharding info every time it's null and a new connection is opened. This way, we get the right _expectedConnectionLength (number of shards x number of connections per shard). The driver naturally maintains the correct number of connections without needing additional logic. We probably need more testing to checks if this is suffisient. WDYT?

EDIT: from the CI you can see that this is flaky, needs some time to establish the right amount of connections

[sylwiaszunejko]

I corrected formatting, but still need support with the things I mentioned here: #8 (comment)

[dkropachev]

@dkropachev the last commit makes the driver update the sharding info every time it's null and a new connection is opened. This way, we get the right _expectedConnectionLength (number of shards x number of connections per shard). The driver naturally maintains the correct number of connections without needing additional logic. We probably need more testing to checks if this is suffisient. WDYT?

EDIT: from the CI you can see that this is flaky, needs some time to establish the right amount of connections

I think it is good enough, as I mentioned on the call there is one possible case is when node is scaled down, we will address it in a separate PR/issue if it is real one.

[sylwiaszunejko]

@dkropachev I added extra checks in the tests, does it look alright? I would also want to clean up the commits, they got a little bit messy, and then we could merge