refactor(message): rewrite parse logic

[olivermrose]

Closes #77

Just need to reimplement automod boundary marking

[Copilot]

Pull Request Overview

This PR refactors the message parsing logic by moving the complex text processing from the UserMessage.toFragments() method to a new dedicated parse.ts module. The main purpose is to improve code organization and maintainability by separating parsing concerns from the message class.

Key changes:

• Replaces the toFragments() method with a nodes getter that uses the new parser
• Introduces a new parsing system with Node types instead of Fragment types
• Updates component interfaces to use the new node structure

Reviewed Changes

Copilot reviewed 8 out of 9 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
src/lib/message/user-message.ts	Removes complex parsing logic and Fragment types, adds nodes getter using new parser
src/lib/message/parse.ts	New parsing module with Node types and parse function for message text processing
src/lib/message/index.ts	Exports the new parse module
src/lib/handlers/eventsub/*	Updates method calls from addAutoModMetadata to setAutoMod
src/lib/components/message/Message.svelte	Updates to use Node types instead of Fragment types
src/lib/components/Emote.svelte	Renames overlays prop to layers
package.json	Adds tldts dependency for URL parsing

Files not reviewed (1)

• pnpm-lock.yaml: Language not supported

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

The URL parsing logic is flawed. The regex /^https?:\/\/|\.$/i will remove trailing dots from all text, not just URLs. For example, 'Hello.' would become 'Hello' and incorrectly be parsed as a URL. Consider checking if the part starts with 'http' before applying URL parsing logic.

Suggested change

	const result = url ? parseTld(url.hostname) : null;
	let url: URL | null = null;
	let result: ReturnType<typeof parseTld> | null = null;
	// Only try to parse as URL if part looks like a URL or domain
	if (
	/^https?:\/\//i.test(part) ||
	// domain-like: contains a dot, no spaces, not just a number
	(/\./.test(part) && !/\s/.test(part) && !/^\d+$/.test(part))
	) {
	const cleaned = part.replace(/^https?:\/\/|\.$/i, "");
	try {
	url = URL.parse(`https://${cleaned}`);
	result = url ? parseTld(url.hostname) : null;
	} catch {
	url = null;
	result = null;
	}
	}

[Copilot]

IRC emotes are being matched by token text rather than by their position ranges. This could cause incorrect matching when the same emote code appears multiple times in a message, as the first occurrence found will be used regardless of position.

Suggested change

	const ircEmote = ircEmotes.find((e) => e.code === part);
	const ircEmote = ircEmotes.find((e) => e.start === start && e.end === end);

[Copilot]

The IRC emote handling creates a node with ircEmote.range.start and ircEmote.range.end positions, but these positions may not align with the current token's position. This could result in overlapping or misaligned nodes when the emote spans multiple tokens or partial tokens.