feat: enhancement for Batcher api to address bulk read

[rahulKQL]

This commit contains enhancements to BatcherAPI.

Currently, there is no ways to know the input user has sent in BatchingDescriptor. There is a good chance that in case of duplicated/non-existing input, the server would only send responses for unique elements.

After this change, the elements sent for batching would be available to the user inside BatchingDescriptor#splitResponse.

[codecov]

Codecov Report

Merging #833 into master will decrease coverage by 0.1%.
The diff coverage is 100%.

@@             Coverage Diff              @@
##             master     #833      +/-   ##
============================================
- Coverage     78.73%   78.62%   -0.11%     
- Complexity     1151     1163      +12     
============================================
  Files           202      203       +1     
  Lines          5107     5142      +35     
  Branches        407      413       +6     
============================================
+ Hits           4021     4043      +22     
- Misses          913      925      +12     
- Partials        173      174       +1

Impacted Files	Coverage Δ	Complexity Δ
...n/java/com/google/api/gax/batching/BatchEntry.java	100% <100%> (ø)	2 <2> (?)
...java/com/google/api/gax/batching/BatcherStats.java	97.4% <100%> (ø)	18 <0> (ø)	⬇️
.../java/com/google/api/gax/batching/BatcherImpl.java	98% <100%> (ø)	16 <0> (ø)	⬇️
...main/java/com/google/api/gax/grpc/ChannelPool.java	45.9% <0%> (-3.04%)	13% <0%> (+2%)
...ain/java/com/google/api/gax/rpc/ClientContext.java	80.88% <0%> (-0.94%)	9% <0%> (+1%)
...gle/api/gax/rpc/InstantiatingWatchdogProvider.java	90.9% <0%> (-0.76%)	16% <0%> (+1%)
...api/gax/grpc/InstantiatingGrpcChannelProvider.java	79.89% <0%> (-0.22%)	35% <0%> (ø)
.../com/google/api/gax/rpc/FixedWatchdogProvider.java	100% <0%> (ø)	10% <0%> (+1%)	⬆️
...src/main/java/com/google/api/gax/rpc/Watchdog.java	80.5% <0%> (+0.91%)	11% <0%> (+5%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update bbefe87...9f42dc9. Read the comment docs.

[igorbernstein2]

I don't love this approach. You are passing 2 parallel lists to splitResponse: elements and batch and you are updating the surface for 2 classes.

I think it would be cleaner for splitResponse to take a list of Entries, which contain the original element and its corresponding unresolved ApiFuture.

[rahulKQL]

Thanks a lot for the suggestion.

I have tried to implement the approach you mentioned. Could you please take a look.

[igorbernstein2]

This is looking good. Can you convert the class to use AutoValue and publish this as a PR?

[igorbernstein2]

This is looking good. @kolea2 can you take a pass as well?

[igorbernstein2]

LGTM except the nit