Optimize JsonCompleter.parse string scanning

[default-anton]

Summary

Optimize JsonCompleter.parse for long streamed string values.

The hot path was Scanners.scan_string. It used to walk ordinary string content one character at a time, even when the next few hundred bytes were just plain text. This change makes it jump to the next interesting character (", \\, or a control character) and copy the whole plain run in one slice.

That keeps the existing parsing behavior, escape handling, and unicode/surrogate validation, but does a lot less Ruby work on the common case.

Before / after

Benchmark command:

JSON_COMPLETER_BENCHMARK=1 bundle exec rspec spec/parse_benchmark_spec.rb

Benchmark settings from the spec run:

• payload bytes: 77690
• prefixes: 9712
• iterations: 50
• chunk size: 8

parse

Metric	Before	After	Change
total runtime	1.9350s	1.2476s	35.5% faster
per iteration	38.700ms	24.952ms	35.5% faster
allocated objects	4,381,588	2,047,569	53.3% fewer

Relative to complete + JSON.parse

Metric	Before	After
speedup	9.07x	12.98x
allocation reduction	4.66x	8.83x

What changed

• added fast-path patterns for string scanning
• appended contiguous plain string runs in slices instead of per character
• kept escape handling, unicode escapes, and surrogate-pair validation on the slower path where it matters
• documented the optimization in README.md
• added a changelog entry

Why this helps

Most streamed JSON text is boring: long runs of normal characters with only occasional escapes or quotes.

Before:

• read one character
• branch on that character
• append one character
• repeat thousands of times

After:

• ask Ruby where the next special character is
• append the whole plain chunk at once
• only switch to character-by-character logic near escapes / terminators

Same result, less interpreter overhead, fewer temporary objects.

Validation

• bundle exec rubocop
• bundle exec rspec
• JSON_COMPLETER_BENCHMARK=1 bundle exec rspec spec/parse_benchmark_spec.rb