xds/clusterimpl: update UpdateClientConnState to handle updates synchronously

[aranjans]

Part of #7210.

UpdateClientConnState now handles updates in a blocking fashion.

To simplify the code, this PR replaces the existing run goroutine inside clusterimpl lb policy and instead handles calls from gRPC and from the child policy (to update state) in a grpcsync.CallbackSerializer. This approach simplifies things by removing the need for synchronization using locks and other grpcsync.Event fields.

RELEASE NOTES: none

[easwars]

I also updated the PR description and also got rid of the release notes since we don't need release notes for non-user visible behavior changes.

[codecov]

Codecov Report

Attention: Patch coverage is 90.62500% with 6 lines in your changes missing coverage. Please review.

Project coverage is 81.82%. Comparing base (b45fc41) to head (fc2054e).
Report is 24 commits behind head on master.

Files with missing lines	Patch %	Lines
xds/internal/balancer/clusterimpl/clusterimpl.go	90.62%	4 Missing and 2 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #7533      +/-   ##
==========================================
- Coverage   82.07%   81.82%   -0.26%     
==========================================
  Files         360      361       +1     
  Lines       27533    27808     +275     
==========================================
+ Hits        22599    22754     +155     
- Misses       3759     3853      +94     
- Partials     1175     1201      +26     

Files with missing lines	Coverage Δ
xds/internal/balancer/clusterimpl/clusterimpl.go	89.57% <90.62%> (+3.16%)	⬆️

... and 51 files with indirect coverage changes

[arvindbr8]

The core modification in this diff appears to be transitioning UpdateClientConnState into a blocking operation. We should explicitly mention this in the commit description.

Apart from a change suggested for a comment block, this diff LGTM.

[easwars]

The core modification in this diff appears to be transitioning UpdateClientConnState into a blocking operation. We should explicitly mention this in the commit description.

UpdateClientConnState always needs to be a blocking operation since it needs to return an error value. And the earlier implementation was not actually doing that correctly.

Yes, we could make that part of the PR description.