Optimize batch read and caching in compactor

[Lujie1996]

Read cache does not cache checkpoint entries. This invalidates the
use of checkpoint batch read in stream layer. This PR adds the extra
parameter CHECKPOINT_READ_BATCH_SIZE to compactor_runner.py
and set it to 1 for compactor.

This PR also enables read cache for compactor to allow just one batch
read of non-checkpoint entries to be cached. This optimizes the read
efficiency of checkpointer.

Related issue: #3651

Checklist (Definition of Done):

• There are no TODOs left in the code
• Coding conventions (e.g. for logging, unit tests) have been followed
• Change is covered by automated tests
• Public API has Javadoc

[codecov]

Codecov Report

Merging #3652 (4b8593d) into master (f2ec6e1) will increase coverage by 0.02%.
The diff coverage is 100.00%.

@@             Coverage Diff              @@
##             master    #3652      +/-   ##
============================================
+ Coverage     77.90%   77.92%   +0.02%     
- Complexity     5042     5043       +1     
============================================
  Files           477      477              
  Lines         23917    23927      +10     
  Branches       2126     2129       +3     
============================================
+ Hits          18632    18644      +12     
+ Misses         4233     4229       -4     
- Partials       1052     1054       +2     

Impacted Files	Coverage Δ
...rc/main/java/org/corfudb/runtime/CorfuRuntime.java	75.32% <100.00%> (ø)
...ava/org/corfudb/runtime/view/AddressSpaceView.java	80.28% <100.00%> (+0.28%)	⬆️

... and 27 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[Maithem]

If the batch can be successfully read and later GC'd/dropped since it's not cached, then that implies that there is enough memory to keep that batch in memory.

Why not set the cache size to the batch size for the compactor? that could improve the compaction performance.

[Lujie1996]

Done.

[SravanthiAshokKumar]

General question - I see that this inefficiency while reading checkpointEntries is common to any client that has to go through the processCheckpoint() method. (Of course the compactor gets affected the most since it has to come up every time with a new runtime)
If that's the case, why to even have the checkpointBatchReadSize default to 5? Can we default it to 1 instead?

[SravanthiAshokKumar]

LGTM!
But changes in CompactorBaseConfig requires changes in CompactorConfigUnitTest

[Lujie1996]

General question - I see that this inefficiency while reading checkpointEntries is common to any client that has to go through the processCheckpoint() method. (Of course the compactor gets affected the most since it has to come up every time with a new runtime) If that's the case, why to even have the checkpointBatchReadSize default to 5? Can we default it to 1 instead?

Good point. I changed the default checkpointReadBatchSize to 1 for corfu runtime.