fix: consider priority_selector when using best_session during rename

[rchl]

Pass point to best_session() so that priority_selector is respected.

I've noticed that having LSP-some-sass and LSP-css running at once sends prepareRename to LSP-some-sass but rename to LSP-css.

Having this fix + prioriority_selector defined in LSP-some-sass will fix the issue.

Related to sublimelsp/LSP-some-sass#4

[jwortmann]

I think the correct way to fix this would be to pass the session name from prepare_provider_session to _on_prepare_result and from there further as an additional argument to when the lsp_symbol_rename command is called a second time. Then if this name is not None, use self.session_by_name instead of self.best_session (in _do_rename) to select the rename session. This would guarantee that it is really the same session.

Otherwise there might be only one of the servers having support for renameProvider.prepareProvider, but another session has a higher priority_selector. In that case, the first session would still be used for the prepare rename request, and then the other session for the rename request.

Passing the point to self.best_session for the renameProvider.prepareProvider is certainly still useful (probably also for various other commands in LSP). I think the code block could be simplified to

point = get_position(self.view, point=args.get('point'))
if not point:
    return None

though I see that in this PR it was just moved from one place to another.

---

Related to #2538

Infinite loop from self reference detected.

[rchl]

Passing the point to self.best_session for the renameProvider.prepareProvider is certainly still useful (probably also for various other commands in LSP). I think the code block could be simplified to

point = get_position(self.view, point=args.get('point'))
if not point:
    return None

Good that someone still remembers what kind of utility functions we have :)

Though in this case I think we should actually not return since the prepare rename request should not need the point. (Not that I ever see point being None here...)

[rchl]

With the addition of session_name for lsp_rename, now the whole fix for the point doesn't even matter for the sublimelsp/LSP-some-sass#4 case anymore. But of course it's still a good fix.

[jwortmann]

Though in this case I think we should actually not return since the prepare rename request should not need the point.

Not sure about that. Isn't the whole point of the prepare rename request to check whether a given location is valid to rename?

The control flow of the rename command is quite complicated, but I think here it probably doesn't matter because it would still return None because "placeholder" is not in args, and then it will eventually return from run if the point is None (which means that there is no cursor in the view).

[jwortmann]

Can be removed