GH-5442 Further optimize multi-threading performance and reduce copy operations in LmdbStore

[kenwenzel]

GitHub issue resolved: #5442

Briefly describe the changes proposed in this PR:

Reduce copy operations in ValueStore and slightly improve multi-threaded insertion perfomace.

---

PR Author Checklist (see the contributor guidelines for more details):

• my pull request is self-contained
• I've added tests for the changes I made
• I've applied code formatting (you can use mvn process-resources to format from the command line)
• I've squashed my commits where necessary
• every commit message starts with the issue number (GH-xxxx) followed by a meaningful description of the change

[hmottestad]

Have you run benchmarks before and after to show how this affects performance?

[kenwenzel]

Yes. I can post something tomorrow.

[kenwenzel]

main:

Benchmark                                                    Mode  Cnt      Score      Error  Units
TransactionsPerSecondBenchmark.largerTransactionLevelNone   thrpt    3     46.978 ±   16.289  ops/s
TransactionsPerSecondBenchmark.mediumTransactionsLevelNone  thrpt    3  23231.246 ± 5066.380  ops/s

Benchmark                        Mode  Cnt    Score     Error  Units
QueryBenchmark.complexQuery      avgt    3   12.399 ±   2.573  ms/op
QueryBenchmark.lots_of_optional  avgt    3  603.818 ± 150.202  ms/op

this PR:

Benchmark                                                    Mode  Cnt      Score      Error  Units
TransactionsPerSecondBenchmark.largerTransactionLevelNone   thrpt    3     48.927 ±   26.386  ops/s
TransactionsPerSecondBenchmark.mediumTransactionsLevelNone  thrpt    3  25608.047 ± 4287.504  ops/s

Benchmark                        Mode  Cnt    Score    Error  Units
QueryBenchmark.complexQuery      avgt    3   11.414 ±  3.412  ms/op
QueryBenchmark.lots_of_optional  avgt    3  610.069 ± 61.084  ms/op

Unfortunately, I can't reliably reproduce benchmark results with my setup (Windows 11 + Cortex XDR + Hyper-V on a notebook).
Its not a huge improvement but tendency is that it should be a bit faster.

[kenwenzel]

@hmottestad I've also added an optimization for Varint that reduces the number of ByteBuffer.put calls.

Benchmark                    Mode  Cnt   Score   Error  Units
QueryBenchmark.complexQuery  avgt    5  10.824 ± 0.607  ms/op