xds/clusterresolver: implement gRFC A61 changes for LOGICAL_DNS clusters

[Pranjali-2501]

Fixes : #8153

This PR implements the "Changes to Logical DNS Clusters" section of gRFC A61 (IPv4/IPv6 Dual-stack Backends).

Currently, LOGICAL_DNS clusters in the xDS cluster resolver have their Load Balancing policy hard-coded to pick_first. This ensures the semantics of connecting to only one address at a time.

This PR updates the xds_cluster_resolver logic. The buildClusterImplConfigForDNS function removes the hard-coded pick_first policy restriction for LOGICAL_DNS clusters, allowing them to use the configured LB policy.
Also, in the DNS update handler, all resolved addresses are now grouped into a single resolver.Endpoint. This ensures that regardless of the configured parent LB policy, the child policy sees a single "backend" endpoint containing all addresses.

RELEASE NOTES:

• xds: LOGICAL_DNS clusters now honor the LB policy configured in the cluster resource, rather than defaulting to a hardcoded pick_first policy.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 83.31%. Comparing base (749af0c) to head (c68ffd6).
⚠️ Report is 4 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #8733   +/-   ##
=======================================
  Coverage   83.30%   83.31%           
=======================================
  Files         419      418    -1     
  Lines       32450    32365   -85     
=======================================
- Hits        27033    26964   -69     
+ Misses       4039     4028   -11     
+ Partials     1378     1373    -5     

Files with missing lines	Coverage Δ
...rnal/xds/balancer/clusterresolver/configbuilder.go	94.16% <100.00%> (+0.04%)	⬆️
.../balancer/clusterresolver/resource_resolver_dns.go	74.19% <100.00%> (-1.20%)	⬇️

... and 45 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[easwars]

This change probably needs a release note.

[easwars]

We want to return a slice that contains a single endpoint.

[easwars]

This is still not doing what we want.

The DNS resolver implementation returns addresses and not endpoints, because the DNS protocol does not know about endpoints. See:

grpc-go/internal/resolver/dns/dns_resolver.go

Line 345 in 4c27cc8

state := resolver.State{Addresses: addrs}

We have code in the channel that converts these addresses to endpoints, with one address per endpoint. See:

grpc-go/resolver_wrapper.go

Line 213 in 4c27cc8

func addressesToEndpoints(addrs []resolver.Address) []resolver.Endpoint {

But the above code will execute only when the DNS resolver is used on the gRPC channel. Here in the cluster_resolver LB policy, we are creating a DNS resolver ourselves, and therefore the code to convert from addresses to endpoints (that exists in the channel) will not be run here. So, the cluster_resolver will see a set of addresses and no endpoints from the DNS resolver in [resource_resolver_dns.go](https://github.com/grpc/grpc-go/pull/8733/files#diff-6249528a41f17a06cec41f598b885840f09ef05ee3740ab297fc5a905f2875ca), and will probably convert it into a single endpoint with all the addresses in it (this is the change that you have in line 138 of resource_resolver_dns.go).

But, at some point, we will change the DNS resolver implementation to do what is being done today in the channel, i.e., return a set of endpoints, with one address per endpoint. That means that this code should handle multiple endpoints as input, but output a single endpoint that contains all the addresses from the input endpoints.

[easwars]

We want to test all possibilities here:

• one endpoint with one address
• one endpoint with multiple addresses
• multiple endpoints, all with one address each
• multiple endpoints, all with multiple addresses.

[easwars]

We also want to make sure we have a more e2e style test to cover this new logic. A LOGICAL_DNS cluster that specifies a round_robin LB policy and the DNS resolver returning multiple addresses. In this case, because cluster_resolver will convert these addresses into a single endpoint, we should end up with all traffic going to the first address (because RR will delegate to PF).