process_messages: don't let a failed error report skip worker teardown

[vtjnash]

A worker should not outlive its master, but the exit(1) enforcing that sat behind error reporting which fails in exactly the circumstance the exit exists for. start_worker redirects a worker's stderr to its stdout, a pipe held only by the master, so when the master dies the worker's EOFError handling writes to a reader-less pipe and raises EPIPE. @error does not contain that: when catch_exceptions(logger) is false it rethrows, and when true the recovery report in
Base.CoreLogging.logging_error writes to the same broken stream unguarded. The throw escaped message_handler_loop's catch, skipping both exit(1) and the stream closes, and the worker then idled in start_worker's wait loop indefinitely, holding its socket in CLOSE-WAIT.

Arm the exit on completion of the task servicing the connection to pid 1 so it no longer depends on anything in the error path succeeding, and guard the remaining raw stderr writes that gate cleanup in deliver_result, connect_to_peer, and the unidentified-peer branch of message_handler_loop.

[codecov]

Codecov Report

❌ Patch coverage is 86.00000% with 7 lines in your changes missing coverage. Please review.
✅ Project coverage is 79.37%. Comparing base (96b42c4) to head (b1cae6f).
⚠️ Report is 2 commits behind head on master.

Files with missing lines	Patch %	Lines
src/process_messages.jl	58.82%	7 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #198      +/-   ##
==========================================
+ Coverage   79.10%   79.37%   +0.26%     
==========================================
  Files          10       10              
  Lines        1977     2027      +50     
==========================================
+ Hits         1564     1609      +45     
- Misses        413      418       +5     

☔ 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.