net: surface errors from setKeepAlive handle calls

[waterWang]

Fixes: #65529

`Socket.prototype.setKeepAlive()` ignored the error code returned by the
underlying handle, so a failed operation (e.g. a negative delay rejected by
the platform with `EINVAL`) reported success and returned the socket. The
adjacent `setTypeOfService()` already throws `ErrnoException` in this case.

This change makes `setKeepAlive()` consistent with `setTypeOfService()`:
when the handle returns a non-zero error and the platform is not Windows
(where keep-alive settings are best-effort), throw the `ErrnoException`.

A regression test stubs the handle to return `UV_EINVAL` and asserts the
error is surfaced with the right `code`/syscall.

[nodejs-github-bot]

Review requested:

• @nodejs/net

[jasnell]

Technically I'm afraid this might be semver-major. I'm happy to consider it just a bug fix if others agree. @nodejs/tsc?

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.12%. Comparing base (54b4e37) to head (68ad950).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #65539      +/-   ##
==========================================
- Coverage   90.13%   90.12%   -0.02%     
==========================================
  Files         751      751              
  Lines      253635   253638       +3     
  Branches    47786    47784       -2     
==========================================
- Hits       228615   228580      -35     
- Misses      16268    16310      +42     
+ Partials     8752     8748       -4     

Files with missing lines	Coverage Δ
lib/net.js	94.56% <100.00%> (+<0.01%)	⬆️

... and 31 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.