More interpreter optimizations

[Saiv46]

Closes #69 and #75

Benchmark results by @Karang (benchmark_all_types):

**ProtoDef-io:master**:
read x 26,817 ops/sec ±1.43% (92 runs sampled)
  ✓ reads (5617ms)
write x 16,188 ops/sec ±0.46% (96 runs sampled)
  ✓ writes (5788ms)
read (compiled) x 104,248 ops/sec ±1.56% (87 runs sampled)
  ✓ reads (compiled) (5474ms)
write (compiled) x 79,582 ops/sec ±0.39% (95 runs sampled)
  ✓ writes (compiled) (5616ms)
**Saiv46:master**:
read x 67,847 ops/sec ±0.45% (95 runs sampled)
  ✓ reads (5676ms)
write x 40,270 ops/sec ±0.38% (95 runs sampled)
  ✓ writes (5558ms)
read (compiled) x 102,333 ops/sec ±1.02% (93 runs sampled)
  ✓ reads (compiled) (5524ms)
write (compiled) x 77,432 ops/sec ±1.04% (86 runs sampled)
  ✓ writes (compiled) (5614ms)

[rom1504]

Can you open a PR for node-minecraft-protocol using the interpreter and this PR as protodef version ?
This would make it possible to check better that this works

[rom1504]

Could you explain what you did here and why it is faster ?

[rom1504]

@Saiv46 do you intend to finish this?

[Saiv46]

I need to figure out what's missing here. Something stopped me from finishing this PR... Oh, yes, fields need to be optimized...

[Saiv46]

Could you explain what you did here and why it is faster ?

@rom1504 First I was just refactoring existing code - first by just splitting it, then applying some optimizations from protodef-neo.

Recently I read this article about optimization fundamentals - so now I know what to do.

By the way:
(UPD. Benchmarks updated, stable ~3x performance improvement)

-  396 passing (1s)
+  396 passing (464ms)

- C:\Users\Admin\node-protodef>npx mocha ./benchmark/benchmark_all_types.js
+ D:\WORKING\Saiv46\node-protodef>npx mocha ./benchmark/benchmark_all_types.js

- read x 8,250 ops/sec ±2.34% (84 runs sampled)
-   √ reads (6102ms)
+ read x 24,708 ops/sec ±7.08% (85 runs sampled)
+   √ reads (5752ms)
- write x 4,229 ops/sec ±12.03% (76 runs sampled)
-   √ writes (5533ms)
+ write x 12,567 ops/sec ±25.72% (81 runs sampled)
+   √ writes (5635ms)

  4 passing (23s)

[Saiv46]

3x interpreter performance improvement achieved!

read x 23,910 ops/sec ±6.52% (80 runs sampled)
  √ reads (6062ms)
write x 12,641 ops/sec ±6.70% (81 runs sampled)
  √ writes (5764ms)

Now I should undo changes to compiler (because this is out of the scope of this PR)
(UPD. Nevermind. Changes revert making compiler slower on benchmarks)

[Saiv46]

I was tried to make something more on this PR, but I run out of ideas that will not break anything.