Remove Send requirement from the lsp

[dsherret]

This is all single threaded, but the lsp code complains if we switch to RefCell and Rc. I tried spawning a local task for this, but that didn't help. Probably some simple fix I don't know about.

deno/cli/npm/resolution/graph.rs

Lines 162 to 163 in 0cce9c2

	// todo(dsherret): for some reason the lsp errors when using an Rc<RefCell<NodeId>> here
	// instead of an Arc<Mutex<NodeId>>. We should investigate and fix.

cc @bartlomieju

[bartlomieju]

This is caused by tower_lsp requiring futures to be Send:

https://docs.rs/tower-lsp/latest/tower_lsp/struct.Server.html#method.serve

It's a low priority, we can revisit it in the future

[dsherret]

I opened ebkalderon/tower-lsp#386

[ebkalderon]

Just left a few comments on that ticket. I would be more than happy to drop the Send bound going forward! As explained in that thread, the required Send bounds have already been dropped in an open WIP branch, but this was done as part of a larger-scale effort that goes beyond the scope of ebkalderon/tower-lsp#386.

If you would find it valuable, I wouldn't mind drafting a new release of tower-lsp which only drops the Send bounds from futures, leaving the remainder of the proposed changes to a future version? The removal of Send bounds is technically orthogonal to everything else that was proposed, so I'd be okay with doing this. 😄