feat(connmgr): lower the default connection pool

[lidel]

This PR aims to close #9420 by lowering the implicit defaults for Swarm.ConnMgr.

Rationale

It will not only help with #9442 but also decrease bitswap gossip.

To apply change to legacy nodes that were initialized before #8913, we need to perform config migration described in #8913 (review)

TODO

• lower high water (50 for now)
• lower low water (150 for now)
• update docs
• create config migration (set Swarm.ConnMgr to {} if values match old defaults)
  • (draft in feat(12-13): removing hardcoded defaults fs-repo-migrations#161)

[Jorropo]

LGTM

[BigLep]

@lidel : I sanity checked the number of Low=50, High=150

By default if a Kubo node has 8GB of total memory, a maxMemory for libp2p will be set to 2GB. That will cause System.ConnsInbound to be 128 I believe (looks like 64 per GB of memory here: https://github.com/libp2p/go-libp2p/blob/master/p2p/host/resource-manager/limit_defaults.go#L345 since Kubo is using those same libp2p defaults: https://github.com/ipfs/kubo/blob/master/core/node/libp2p/rcmgr_defaults.go#L94). As a result, in this particular setup, our high water mark may not get hit (unless there are sufficient outbound connections). I know there isn't a perfect number here and we have no stats on the systems that folks are using. I would be in favor of dropping the ConnMgr.High to 120 instead.

[lidel]

Continued in #9483