Preserve original fd blocking state in TLS I/O operations

[xbasel]

This change prevents unintended side effects on connection state and improves consistency with non-TLS sync operations.

For example, when invoking connTLSSyncRead with a blocking file descriptor, the mode is switched to non-blocking upon connTLSSyncRead exit. If the code assumes the file descriptor remains blocking and calls the normal read expecting it to block, it may result in a short read.

This caused a crash in dual-channel, which was fixed in this PR by relocating connBlock():
#837

[codecov]

Codecov Report

Attention: Patch coverage is 44.44444% with 5 lines in your changes missing coverage. Please review.

Project coverage is 70.70%. Comparing base (2df56d8) to head (aa0db5c).
Report is 21 commits behind head on unstable.

Files with missing lines	Patch %	Lines
src/anet.c	44.44%	5 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##           unstable    #1298      +/-   ##
============================================
+ Coverage     70.69%   70.70%   +0.01%     
============================================
  Files           114      115       +1     
  Lines         63161    63185      +24     
============================================
+ Hits          44650    44678      +28     
+ Misses        18511    18507       -4     

Files with missing lines	Coverage Δ
src/tls.c	100.00% <ø> (ø)
src/anet.c	73.23% <44.44%> (-0.87%)	⬇️

... and 26 files with indirect coverage changes

---- 🚨 Try these New Features:

• Flaky Tests Detection - Detect and resolve failed and flaky tests
• JS Bundle Analysis - Avoid shipping oversized bundles

[xbasel]

Please note that timeouts are still not restored. Not sure we should worry about timeouts.

[ranshid]

I agree that the timeout is not the thing we should keep track of, however I wonder why we cannot just flag the connection with indication that it is blocked instead of going through the fcntl all the time?

[xbasel]

I agree that the timeout is not the thing we should keep track of, however I wonder why we cannot just flag the connection with indication that it is blocked instead of going through the fcntl all the time?

Using an app-level flag adds complexity and risks desync with the kernel. fcntl is lightweight, reliable, and avoids duplication.

[ranshid]

@xbasel can you please signoff your commit?

Also please elaborate more on the top comment. provide some example etc...