xds: resubmit xds client pool changes from #7898 along with fix to set fallback bootstrap config from googledirectpath to xdsclient pool

[purnesh42H]

Change #7898 introduced a pool of xDS clients that share the same configuration within the pool, instead of using a single global map of xDS clients.

After #7898 changes, production path of xdsclient will maintain a default pool populated with the bootstrap configuration during init() using the internal bootstrap.GetConfiguration() function. This function reads the configuration from environment variables. Currently, bootstrap.GetConfiguration() also includes logic to read the fallback configuration set by the c2pResolver (from the googledirectpath library) during its Build() method. After change #7898, this Build() method executes after init(), so the fallback bootstrap configuration set by the c2pResolver is not assigned to xdsclient.DefaultPool.

#7898 was reverted because c2p resolver was broken. This pull request resubmit changes from #7898 and modifies the c2pResolver to set the fallback bootstrap configuration on the assigned xDS client pool (defaulting to xdsclient.DefaultPool) and removes the fallback bootstrap configuration logic from internal/xds/bootstrap.

RELEASE NOTES:

• xds: Implement an xDS client pool to allow creating multiple servers/channels with varying bootstrap configurations within a single test.

[codecov]

Codecov Report

Attention: Patch coverage is 80.70866% with 49 lines in your changes missing coverage. Please review.

Project coverage is 82.18%. Comparing base (1318104) to head (0ebd548).
Report is 6 commits behind head on master.

Files with missing lines	Patch %	Lines
xds/internal/xdsclient/pool.go	78.78%	15 Missing and 6 partials ⚠️
xds/internal/xdsclient/clientimpl.go	84.48%	13 Missing and 5 partials ⚠️
internal/testutils/xds/e2e/setup/setup.go	16.66%	3 Missing and 2 partials ⚠️
xds/internal/resolver/xds_resolver.go	72.72%	2 Missing and 1 partial ⚠️
xds/googledirectpath/googlec2p.go	50.00%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #8050      +/-   ##
==========================================
- Coverage   82.23%   82.18%   -0.06%     
==========================================
  Files         387      387              
  Lines       39015    39042      +27     
==========================================
+ Hits        32085    32086       +1     
- Misses       5591     5619      +28     
+ Partials     1339     1337       -2     

Files with missing lines	Coverage Δ
internal/xds/bootstrap/bootstrap.go	65.91% <100.00%> (-2.95%)	⬇️
xds/internal/xdsclient/client.go	100.00% <100.00%> (ø)
xds/internal/xdsclient/xdsresource/errors.go	100.00% <ø> (ø)
xds/server.go	82.20% <100.00%> (-0.84%)	⬇️
xds/server_options.go	100.00% <100.00%> (ø)
xds/googledirectpath/googlec2p.go	93.33% <50.00%> (+0.22%)	⬆️
xds/internal/resolver/xds_resolver.go	78.65% <72.72%> (-0.73%)	⬇️
internal/testutils/xds/e2e/setup/setup.go	64.70% <16.66%> (-17.65%)	⬇️
xds/internal/xdsclient/clientimpl.go	79.16% <84.48%> (+6.21%)	⬆️
xds/internal/xdsclient/pool.go	78.78% <78.78%> (ø)

... and 24 files with indirect coverage changes

[purnesh42H]

@purnesh42H : I discussed offline with @dfawley and we are in agreement that just my review is enough for this PR. So, please address the remaining comments and the merge conflicts and assign it back to me when its good to be looked at. Thanks.

Thanks. I have added back the xds client pool changes and resolved conflicts. Its ready for review.

[easwars]

Nit: s/defaultStreamBackoffFunc/defaultExponentialBackoff

We are trying to following a style where we name the variable which stores the function being overridden with the same name as the latter. So, something like a rand.Shuffle being overridden would have a name of randShuffle or randShuffleFunc so that in non-test code when you see the variable name, you kind of know what the actual function is.

[easwars]

We don't run any tests of the same package in parallel. A grep for t.Parallel returns no results in our repo.

easwars@minetop: grpc-go$ grep -r "t.Parallel" *
easwars@minetop: grpc-go$ 

Tests from different packages are automatically run in parallel by the go test command though. So, if you are seeing problems with this, then some other tests also use the global default pool. Can you please check? Thanks.