Replace conditional logic in add(), subtract() with bit manipulation

[sgup432]

Description

I came across this logic in add(), subtract() in NumericUtils where we have branching logic to extract carry/borrow.
Seems like we can better replace it with bitwise logic, and it would be faster.

[github-actions]

This PR does not have an entry in lucene/CHANGES.txt. Consider adding one. If the PR doesn't need a changelog entry, then add the skip-changelog label to it and you will stop receiving this reminder on future updates to the PR.

[kaivalnp]

I tried benchmarking your proposed changes, and looks like subtract is slowed down, while add is marginally faster:

Benchmark                          (size)   Mode  Cnt   Score   Error   Units
NumericUtilsBenchmark.add              16  thrpt   15  68.679 ± 2.212  ops/us
NumericUtilsBenchmark.add              32  thrpt   15  37.675 ± 0.392  ops/us
NumericUtilsBenchmark.add              64  thrpt   15  19.586 ± 0.233  ops/us
NumericUtilsBenchmark.addNew           16  thrpt   15  70.461 ± 0.161  ops/us
NumericUtilsBenchmark.addNew           32  thrpt   15  38.566 ± 0.035  ops/us
NumericUtilsBenchmark.addNew           64  thrpt   15  20.022 ± 0.019  ops/us
NumericUtilsBenchmark.subtract         16  thrpt   15  68.377 ± 2.006  ops/us
NumericUtilsBenchmark.subtract         32  thrpt   15  35.782 ± 0.995  ops/us
NumericUtilsBenchmark.subtract         64  thrpt   15  19.097 ± 0.393  ops/us
NumericUtilsBenchmark.subtractNew      16  thrpt   15  51.872 ± 0.097  ops/us
NumericUtilsBenchmark.subtractNew      32  thrpt   15  27.446 ± 0.012  ops/us
NumericUtilsBenchmark.subtractNew      64  thrpt   15  14.163 ± 0.009  ops/us

See #15303 for my benchmark setup, where I have another proposal to speed these up!

[jainankitk]

Guess we can close this PR in favor of ? - #15303