Prevent errors from propagating to spawned

[ElFantasma]

spawned catches errors on its callbacks (like init, handle_call and handle_cast) and logs them. If a recurrent error happens, it will spam the logs.
A proper callback implementation should avoid propagating these errors and decide instead what to do (ie. log them or ignore them or fail)
In particular, init on RLPxConnection is currently spamming a lot of

2025-07-03T15:29:06.625135Z ERROR spawned_concurrency::tasks::gen_server: Initialization failed: IoError(Os { code: 54, kind: ConnectionReset, message: "Connection reset by peer" })

A temporary fix was issued, but we should catch the error in ethrex and handle it instead.

Once this is fixed, reset the log filter to allow showing other potential errors on callbacks.