make the source hyperlinks work in the fuzzing web interface (currently they construct autodoc URLs)

[andrewrk]

Extracted from #20958.

[nix-shell:~/dev/zig/build-release]$ stage3/bin/zig build test-std  -Dtest-filter="tokenizer" -Dskip-release -Dskip-non-native -Dskip-libc -Dskip-single-threaded --fuzz --debug-rt --port 41099

Text from screenshot reproduced below:

fn testTokenize(source: [:0]const u8, expected_token_tags: []const Token.Tag) !void {
    var tokenizer = Tokenizer.init(source);
    for (expected_token_tags) |expected_token_tag| {
        const token = tokenizer.next();
        try std.testing.expectEqual(expected_token_tag, token.tag);
    }
    // Last token should always be eof, even when the last token was invalid,
    // in which case the tokenizer is in an invalid state, which can only be
    // recovered by opinionated means outside the scope of this implementation.
    const last_token = tokenizer.next();
    try std.testing.expectEqual(Token.Tag.eof, last_token.tag);
    try std.testing.expectEqual(source.len, last_token.loc.start);
    try std.testing.expectEqual(source.len, last_token.loc.end);
}

For example in the above snippet, the expectEqual link points to http://127.0.0.1:41099/#.home.andy.dev.zig.lib.std.std.testing.expectEqual.

Instead, it should point to one of these:

• #std.testing.expectEqual
• #f123d456 where 123 and 456 are the file and decl indexes respectively.
• #/home/andy/dev/zig/lib/std/zig/tokenizer.zig:123:456 where 123 and 456 are line and column respectively.

And then the onHashChange logic needs to respond by loading the appropriate file and scrolling the location into view.