fix(share/shwap/p2p/shrex/peers): disconnect peers that fail every request

[renaynay]

When a peer is put on cooldown, it can re-enter peer rotation almost immediately (3s) after they were put on cooldown, and with a 5-peer pool, this could happen frequently (wasting requests on an less responsive peer).

This PR implements a solution to count consecutive failed requests per peer and, at peerStrikeLimit, remove the peer from the nodes pool and close the connection instead of cooling it down.

Rules:

• A successful request clears the count, so only peers that fail every request are kicked.
• The disconnect makes discovery discard the peer and back off from redialing it for 10 minutes, while its 1s loop looks for a replacement, so the peer stays out of rotation for minutes rather than
  seconds.
• Strikes are held in a bounded LRU as peers churn.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 69.23077% with 12 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (feature/shrex-eds-optimisation@00e2029). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
share/shwap/p2p/shrex/peers/metrics.go	27.27%	8 Missing ⚠️
share/shwap/p2p/shrex/peers/manager.go	85.18%	2 Missing and 2 partials ⚠️

Additional details and impacted files

@@                        Coverage Diff                        @@
##             feature/shrex-eds-optimisation    #5146   +/-   ##
=================================================================
  Coverage                                  ?   37.26%           
=================================================================
  Files                                     ?      311           
  Lines                                     ?    21626           
  Branches                                  ?        0           
=================================================================
  Hits                                      ?     8058           
  Misses                                    ?    12555           
  Partials                                  ?     1013           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[renaynay]

Concerns with this PR:

• With ConcurrencyLimit 16, several concurrent requests to one peer can each fail and each observe a count at or above the limit before resetStrikes inside kickPeer lands. Every one calls kickPeer.
• Kick closes a connection the whole node shares host-wide
• Strikes count things that are not just peer faults
• Strikes can empty out peer pool - and the archival node pool is not healthy enough on any network to combat that risk

Re-running archival benchmarks without this commit to see if had demonstrable impact on the shrex-eds speed-up