Speed up JSON string serialization

[jonathan343]

Summary

Replaces our custom JSON string escaping code with Python's standard JSON string encoder, json.encoder.encode_basestring.

The JSON bytes stay the same for valid strings, including non-English text and emoji. Strings that cannot be encoded as UTF-8 still raise an error. They now fail before writing any part of that string, and the error details may change.

Performance testing

Compared develop with this branch on Python 3.12.14 using single-string inputs of different sizes. Each input is ASCII, so each input character is one byte before JSON escaping.

The old code makes one replacement for each double quote, backslash, or control character from U+0000 through U+001F. The test inputs contain an evenly spread mix of quotes, backslashes, newlines, tabs, and NUL characters, with roughly 0%, 1%, or 10% of characters needing escaping. The counts below match the original regex exactly.

Times are medians of 33 samples per case, measuring the full JSONCodec.serialize call. Input construction and correctness checks are outside the timed portion.

Input bytes	Characters needing escaping	Before	After	Speedup
1,024	0	2.7 µs	2.0 µs	1.4x
1,024	10	3.9 µs	2.2 µs	1.7x
1,024	102	13.2 µs	2.3 µs	5.8x
16,384	0	33.8 µs	24.1 µs	1.4x
16,384	163	50.9 µs	25.0 µs	2.0x
16,384	1,638	188.8 µs	25.7 µs	7.3x
262,144	0	530.7 µs	390.6 µs	1.4x
262,144	2,621	795.6 µs	393.3 µs	2.0x
262,144	26,214	3053.2 µs	404.9 µs	7.5x

Both versions produced identical JSON bytes for every input. These are synthetic string benchmarks, not measurements of AWS request latency.

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.