`Wrapper#accept` might be crashing in production.

[ioquatix]

io-endpoint/lib/io/endpoint/wrapper.rb

Line 157 in e82badc

socket, address = server.accept

    1m     warn: Async::Task [oid=0xdfc] [ec=0xde8] [pid=56009] [2024-05-08 20:40:28 +1200]
               | Task may have ended with unhandled exception.
               |   OpenSSL::SSL::SSLError: SSL_accept returned=1 errno=0 peeraddr=[::1]:50598 state=error: ssl/tls alert certificate unknown
               |   → /home/samuel/.rubies/ruby-3.3.1/lib/ruby/3.3.0/openssl/ssl.rb:527 in `accept'
               |     /home/samuel/.rubies/ruby-3.3.1/lib/ruby/3.3.0/openssl/ssl.rb:527 in `accept'
               |     /home/samuel/.gem/ruby/3.3.1/gems/io-endpoint-0.10.2/lib/io/endpoint/wrapper.rb:157 in `accept'
               |     /home/samuel/.gem/ruby/3.3.1/gems/io-endpoint-0.10.2/lib/io/endpoint/generic.rb:83 in `block in accept'
               |     /home/samuel/.gem/ruby/3.3.1/gems/io-endpoint-0.10.2/lib/io/endpoint/bound_endpoint.rb:65 in `block (2 levels) in bind'
               |     /home/samuel/.gem/ruby/3.3.1/gems/async-2.11.0/lib/async/task.rb:164 in `block in run'
               |     /home/samuel/.gem/ruby/3.3.1/gems/async-2.11.0/lib/async/task.rb:377 in `block in schedule'

1. Need to investigate whether SSLSocket#accept is performing additional TLS negotiation (is it causing latency in the accept loop?)
2. Need to figure out the best way to rescue errors to avoid killing the accept loop.

[ioquatix]

#10