Feature request: Add error_handler_t support to json::parse() for tolerant handling of invalid UTF-8 in input strings

[aaronalbers]

🚀 Feature Request: Tolerate / replace invalid UTF-8 during parsing

Current behavior

The library strictly enforces that input JSON text is valid UTF-8 (as per RFC 8259). When parsing a string value that contains invalid UTF-8 byte sequences, json::parse() throws a json::parse_error (usually code 101: "invalid string: ill-formed UTF-8 byte").

This is correct per the JSON specification, but in real-world applications we often need to process "dirty" JSON from untrusted sources (user input, legacy systems, corrupted network data, logs, scraped web content, etc.) where strings may contain invalid sequences.

Desired feature

Add an optional error_handler_t parameter (same enum already used in dump()) to the main json::parse() overloads, allowing callers to choose:

• strict (current/default): throw on invalid UTF-8 (existing behavior)
• replace: replace invalid sequences with U+FFFD (�) and continue parsing
• ignore: skip invalid bytes entirely and continue parsing

Example proposed API (mirroring dump() style):

static basic_json parse(
    detail::input_adapter&& i,
    const parser_callback_t cb = nullptr,
    const bool allow_exceptions = true,
    const bool ignore_comments = false,
    error_handler_t error_handler = error_handler_t::strict  // new optional param
);