Improve BinaryFieldMapper.CustomBinaryDocValuesField.bytes()

[nknize]

Falling out of #6884 this issue is to explore improving (and benchmarking) the BinaryFieldMapper.CustomBinaryDocValuesField.bytes() method to prevent sorting and deduping on every call. Options include switch from an ArrayList<byte[]> to a TreeSet<byte[]> or using sort/dedup after every add.

[nknize]

@navneet1v can I assign this one to you so you can own this exploration either here or upstream in Lucene? No rush... and I'd look at this as a follow on to just getting MultiValue in Lucene. I don't expect any performance improvements whether it's a TreeSet or something else to be terribly impactful but every little bit helps.

[navneet1v]

@navneet1v can I assign this one to you so you can own this exploration either here or upstream in Lucene? No rush... and I'd look at this as a follow on to just getting MultiValue in Lucene. I don't expect any performance improvements whether it's a TreeSet or something else to be terribly impactful but every little bit helps.

Yes sure.

[nknize]

Excellent!! Thank you!

[kkmr]

I chatted with @navneet1v. I will be picking up this issue.

[navneet1v]

I chatted with @navneet1v. I will be picking up this issue.

Thanks @kkmr for picking this up.