Avoid scanning the entire input to detect parseInt hexadecimal prefixes #198

hxParseIntChars in src/hx/StdLibs.cpp calls is_hex_string(str, strlen(str)) after skipping whitespace. is_hex_string only checks the first few characters and whether the string is at least three or four characters long. Computing the full length scans an otherwise irrelevant suffix before integer parsing starts.

I reproduced this against checkout 03a4e8d on Linux x86_64 through WSL, Haxe 4.3.3, HXCPP_M64 and hxcpp_smart_strings. A native probe called __hxcpp_parse_int 500 times with preconstructed byte strings:

• "7x": 0.006 ms, checksum 3500.
• "7x" followed by 1048576 a characters: 3.768 ms, checksum 3500.

Both inputs stop parsing at x and return 7. Construction was outside the timed region. These single-run diagnostic measurements show unnecessary suffix-dependent work, not an estimated speedup for ordinary short numeric strings.

Detect the optional sign and hexadecimal prefix with bounded checks instead of strlen. Ensure each character exists before inspecting the next one; empty strings, a sign alone, "0", "0x" and signed incomplete prefixes must retain their current behavior. Avoid replacing a safe length check with an out-of-bounds short-string read.

The same helper serves substring integer parsing, so preserve that API's bounds and terminator behavior. Benchmark normal decimal numbers, hex, whitespace, short invalid inputs and short numeric prefixes with long suffixes. Keep overflow and saturation behavior unchanged.

Related: #151 for the performance suite. This specifically removes a full-input scan that is unnecessary for prefix detection.