added performance boost to convert case

[alechirsch]

I ran the benchmark before and after the changes. It seems to have made deserialization MUCH faster, and a smaller performance boost to serialize.
Resolves #84

Before convertCase optimization

Platform info:

Darwin 18.6.0 x64
Node.JS: 10.15.1
V8: 6.8.275.32-node.12
Intel(R) Core(TM) i5-5287U CPU @ 2.90GHz × 4

Suite:

serializeAsync x 71,616 ops/sec ±1.76% (75 runs sampled)
serialize x 110,926 ops/sec ±4.38% (81 runs sampled)
serializeConvertCase x 70,255 ops/sec ±2.48% (91 runs sampled)
deserializeAsync x 152,227 ops/sec ±2.29% (80 runs sampled)
deserialize x 394,181 ops/sec ±1.24% (91 runs sampled)
deserializeConvertCase x 30,294 ops/sec ±1.01% (92 runs sampled)
serializeError x 284,856 ops/sec ±1.06% (92 runs sampled)
serializeError with a JSON API error object x 12,753,945 ops/sec ±4.90% (91 runs sampled)

After convertCase optimization

Platform info:

Darwin 18.6.0 x64
Node.JS: 10.15.1
V8: 6.8.275.32-node.12
Intel(R) Core(TM) i5-5287U CPU @ 2.90GHz × 4

Suite:

serializeAsync x 70,198 ops/sec ±3.08% (76 runs sampled)
serialize x 119,645 ops/sec ±3.64% (85 runs sampled)
serializeConvertCase x 91,489 ops/sec ±3.21% (88 runs sampled)
deserializeAsync x 144,878 ops/sec ±5.55% (70 runs sampled)
deserialize x 361,896 ops/sec ±3.33% (87 runs sampled)
deserializeConvertCase x 133,200 ops/sec ±1.91% (91 runs sampled)
serializeError x 281,794 ops/sec ±1.05% (84 runs sampled)
serializeError with a JSON API error object x 12,884,729 ops/sec ±1.89% (86 runs sampled)

[coveralls]

Coverage remained the same at 100.0% when pulling 9492711 on alechirsch:convert-case-performance into a32ba8d on danivek:master.

[alechirsch]

Added LRU cache, here are the new benchmarks

serializeAsync x 62,434 ops/sec ±13.04% (69 runs sampled)
serialize x 125,513 ops/sec ±2.49% (81 runs sampled)
serializeConvertCase x 91,754 ops/sec ±2.83% (88 runs sampled)
deserializeAsync x 161,124 ops/sec ±3.41% (80 runs sampled)
deserialize x 392,788 ops/sec ±1.19% (89 runs sampled)
deserializeConvertCase x 114,102 ops/sec ±0.55% (88 runs sampled)
serializeError x 267,726 ops/sec ±3.77% (83 runs sampled)
serializeError with a JSON API error object x 12,156,898 ops/sec ±3.78% (85 runs sampled)

[danivek]

Awesome! I think it's the best way to avoid memory leaks. Good job 👍

[alechirsch]

@danivek Do you need anything else from me?