[Refactor] Using XtraMCP parsing

[wjiayis]

I was reading XtraMCP more thoroughly just now, and I realized that it parses bibliography more rigorously than what's implemented in PaperDebugger (for citation suggestion) -- which has fewer edge case handling.

The refactor would also help with abstraction of tasks, where document parsing is allocated to XtraMCP.

A downside would be that the citation suggestion function would not be usable if XtraMCP is not connected (eg in local development) -- perhaps have the current strategy as a fallback?

I'm thinking of refactoring PaperDebugger's bibliography to read BibEntry from XtraMCP and take parsed fields, what do you think @Junyi-99 @4ndrelim ? If it makes sense, I'll swap it out next weekend.

[4ndrelim]

Hi @wjiayis, yup im ok with it. Just worried if it might introduce alot of latency? If not, we can centralise and use the logic from XtraMCP's backend.