feat(lsp): completion side effects

[MariaSolOs]

Closes #25714

This PR introduces vim.lsp.completion with the goal of supporting the completion behaviours (snippet expansion, execution of commands, application of text edits) specified by LSP.

To enable completions for a given buffer and client, users should call vim.lsp.completion.enable(true, client_id, bufnr, opts). Currently the options parameter includes the autotrigger property, which when set will request completions after a trigger character has been typed.

This PR also sets new default keymaps for the snippet navigation:

• <Tab> will jump to the next tabstop if the snippet is active and jumpable forwards.
• <S-Tab> will jump to the previous tabstop if the snippet is active and jumpable backwards.

If these keys are being used in other keymaps, they will be restored once the snippet session ends.

[echasnovski]

This function is mostly called in "scheduled" fashion (via vim.shedule() or after vim.schedule_wrap()). Usually it should be possible to do single schedule_wrap() after defining function instead of all other scheduling in place.

[MariaSolOs]

Okay the implementation is done and tests are all passing. At this point we just need to agree on whether we should enable completion by default and the structure of the public API (so just the usual bikeshedding ;)).

Not sure if we want to do that in this PR or in a separate one. Wdyt @mfussenegger @clason @justinmk @gpanders?

[justinmk]

not a blocker: did you intend to revert your supports_method changes in lsp.txt ?

[MariaSolOs]

not a blocker: did you intend to revert your supports_method changes in lsp.txt ?

Yeah I didn't want to keep the changes in that file with stuff unrelated to this PR. I plan on doing a better review of the current LSP docs later.