Fixed feign-micrometer exception handling.

[r-dimitrov]

The current MR improves exception handling for feign-micrometer clients. Previously, exceptions such as 404, 400, 500, etc., were not marked with error=true in the Observation and thus were not reported in the trace as errors.

The exception declaration has been changed from (FeignException ex) to (Exception ex), as none of the supported clients (e.g., Apache HttpClient, OkHttp) throws FeignException, but rather client-specific exceptions.

The fix is currently in production in my organization, and it has been working effectively.

Fixes #2566

[r-dimitrov]

I've just added some fixes, as there were exceptions when some clients had defined a custom ErrorDecoder. The input stream is consumed, and the decoder throws an exception. This way, we are copying the input stream so it can be used multiple times.

[kdavisk6]

Can we discuss the need to read the request body into memory more?

[kdavisk6]

See my other comments

[r-dimitrov]

Hello, @velo. Why did you close the PR :( Lets find a solution and make it work..

[velo]

Ops, closed by mistake, my bad

[r-dimitrov]

@velo @kdavisk6 Hello, guys! Could we proceed with merging the PR? It's in good shape, and it's truly a must-have feature when it comes to distributed tracing. Thank you so much for all the great work you do — it's truly appreciated!

[diniodinev]

This PR addresses a gap in Micrometer tracing for Feign clients where exceptions such as 404, 400, 500 were not being marked with error=true in the Observation. As a result, these errors were not visible in distributed traces, making it harder to monitor and troubleshoot failed external calls.
So we need that also in our organization. Any updates when it will be ready?