Refresh diagnostics on open, on close, on change, and after R execution

[DavisVaughan]

Addresses posit-dev/positron#2155
Part of posit-dev/positron#1005

Okay, after much fiddling, I've come up with the following Alpha level scheme for our diagnostics:

• did_open(): Runs diagnostics for that file
• did_change(): Runs diagnostics for that file if we are up to date on changes
• did_close(): Clears diagnostics for that file
• did R console execution: Runs diagnostics for all open files

There is much more we can eventually do, like running diagnostics for the whole workspace, maintaining diagnostics on close (and updating them through a "related documents" feature), etc, but this should be a pretty nice improvement.

I'd also like to make a follow up PR or two that trims down our usage of R while diagnostics are being run. We currently look up the set of installed packages and environment symbols at each diagnostic call, but really I think we can cache these and only update them when we do a diagnostic refresh after R console execution (and on startup).

A few of the high level changes:

• Replaced lsp_client with the overarching lsp_backend in RMain, because the backend holds the list of the currently open documents, and we need that to refresh them after an R console execution
• You'll see an IndexerStateManager, this is solely used to ensure that we don't run any diagnostics until the indexer has had a chance to run (30 second timeout). The indexer is a very important part of our diagnostics, and we can return some pretty bad ones if we try to run diagnostics before the indexer has finished. Since the indexer runs in a tokio thread, this is totally possible and does happen for me in dplyr where it takes around 3 second for the indexer to fully run. I think I've managed to make the initialization check pretty cheap.
• After handle_execute_request() (R execution), we refresh diagnostics in all open documents by requesting an r_async_task() that loops over the open documents and spawns a task on the LSP tokio runtime for each of them, requesting a diagnostic refresh.

Note: We are still undecided on posit-dev/positron#1325, which is about what we should do when you are working on an R package and haven't run load_all() yet, or if you have functions in a script with a library(dplyr) call at the top but you haven't run it yet. But the big benefit of this PR is that when you do run them, the diagnostics automatically update.

Untitled.mov

[kevinushey]

LGTM!