collator-protocol: short-term fixes for connectivity

[ordian]

Currently, collators add validators to the reserved peerset when a collation is ready and only update it on the next collation.
That means that after a validator disconnects a collator, it will still try to connect back, even if it's not in their backing group.

As a short term fix, we could:

1. Issue a connection request (add to the reserved peerset) on every relay parent activation. This will ensure that we change groups on time on the collator side as well as pre-connection. The downside is that everyone will try to occupy a slot now.
2. Increase the collation peer slots to 50 or 100 to counter the negative effects of 1. Currently, parachain teams have that many collators per parachain.
3. Reduce the

   polkadot/node/network/collator-protocol/src/lib.rs

   Line 61 in c87a220

   inactive_collator: Duration::from_secs(24),

   to 6 or 5 seconds.

Originally posted by @ordian in #4381 (comment)