transport_service: Improve connection stability by downgrading connections on substream inactivity

[lexnv]

This PR advances the keep-alive timeout of the transport service.

Previously, the keep-alive timeout was triggered 5 seconds after the connection was reported to the transport service regardless of substream activity.

• (0secs) T0: connection established; keep-alive timeout set to 5seconds in the future
• (4secs) T1: substream A, B, C opened
• (5secs) T2: keep-alive timeout triggered and the connection is downgraded. T1 was not taken into account, otherwise, the keep-alive timeout should be triggered at second 9 (T1 at 4 seconds + keepalive 5 seconds)
• (6secs) T3: substreams A, B, C closed -> connection closes
• (7secs) T4: cannot open new substreams even if we expected the connection to be kept alive for longer

In this PR:

• KeepAliveTracker structure to forward the keep-alive timeout of connections.
• Connection ID is forwarded to SubstreamOpened events to identify properly substream Ids. This is needed because the ConnectionContext contains up to two connections (primary and secondary)

Testing Done

• test to ensure keepalive downgrades the connection after 5 seconds
• test to ensure keepalive is forwarded on substream activity
• test to ensure a downgraded connection with dropped substreams is closed

Closes #253.

[lexnv]

Will have another pass over KeepAliveTracker. Im not entirely happy at the moment that we need a service notification to advance the polling here, will have to see how this plays with low-connectivity environments

[lexnv]

A simplified example in the rust playground where I've reproduced this: https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=182bf56449de7827c8f83b649dbba4fb