feat: add support for RegExp.leftContext RegExp.rightContext

[ammarahm-ed]

This pull request to quickjs.c introduces several changes to enhance the handling of regular expression contexts within the JSContext structure. The most important changes include adding new fields to store RegExp context information, initializing these fields in context creation, freeing them in context destruction, and implementing new functions to manage these contexts.

Enhancements to RegExp context handling:

• quickjs.c: Added new fields regexp_left_ctx, regexp_right_ctx, regexp_last_match_str, regexp_last_match_start, and regexp_last_match_end to the JSContext structure to store RegExp context information.
• quickjs.c: Initialized the new RegExp context fields in the JS_NewContextRaw function.
• quickjs.c: Freed the new RegExp context fields in the JS_FreeContext function to prevent memory leaks.
• quickjs.c: Implemented new functions js_regexp_get_leftContext and js_regexp_get_rightContext to retrieve and manage the left and right context of the last RegExp match.
• quickjs.c: Updated the js_regexp_exec function to store the last match string and its start and end positions in the new context fields.

// Define a string to test
const testString = "Hello, world!";

// Create a regular expression to match the word "world"
const regex = /world/;

// Execute the regular expression on the test string
const result = regex.exec(testString);

// Check if the match was successful
if (result) {
    // Print the matched result
    console.log("Matched:", result[0]);

    // Print the left context
    console.log("Left Context:", RegExp.leftContext);

    // Print the right context
    console.log("Right Context:", RegExp.rightContext);
} else {
    console.log("No match found.");
}

Closes #921

[bnoordhuis]

One problem I see with this approach is that regexp_last_match_str can stay around indefinitely (also applies to the substrings.)

Imagine you run a regex on a 100 MB string. That string can't be reclaimed until the next time a regexp runs - which may be never.

[bnoordhuis]

You need to mark these in JS_MarkContext, otherwise the GC might free them prematurely.

[bnoordhuis]

Suggested change

	JS_FreeValue(ctx, ctx->regexp_left_ctx);
	JS_FreeValue(ctx, ctx->regexp_left_ctx);
	ctx->regexp_left_ctx = JS_UNDEFINED;

Avoids double frees, etc. Applies to the other JS_FreeValue calls below too.

[bnoordhuis]

We don't have a strict line length but try to keep it under 80 columns whenever you can, by doing something like:

JSString *p;
// ...
p = JS_VALUE_GET_STRING(ctx->regexp_last_match_str)
ctx->regexp_left_ctx = js_sub_string(ctx, p, 0, ctx->regexp_last_match_start);

Also, bad things will happen when ctx->regexp_last_match_str is not a string. I realize that's guarded by the ctx->regexp_last_match_start >= 0 check but it seems kind of fragile. Why not:

if (JS_IsString(ctx->regexp_last_match_str)) {
    // ...
}