Provide completions in the "value" position

[jennybc]

This is a variation / follow-up on #770.

We generally don't provide completions in the "value" position, in the sense of fn(name = value). This feels weird to me. It's admittedly a somewhat pedantic point, but I don't think it's pure pedantry. It feels like this violates the logic of how completions should work.

Pro tip: An explicit gesture to get completions is Ctrl/Cmd + Space. This works in some contexts where, say, pressing "tab" inserts a literal \t vs. triggering completions.

Consider this code:

a_kinda_long_name <- letters[1:3]
another_long_name <- letters[4:6]

append(x = a_kinda_long_name, y = another_long_name)

Below I indicate cursor position with @.

If you explicitly ask for completions at x = @, you get 'No suggestions'.

But as soon as you type anything, e.g. x = a@, you get dozens of completions, including a_kinda_long_name and another_long_name.

It feels like the implicit contract is that the completions for x = an@ are a subset of the completions for x = a@ are a subset of the completions for x =@. And we're currently violating that.

You can actually see that in this recording, where I request completions before I've typed anything, then get stuck in the land of 'No suggestions' until I backspace all the way back to the "=":

Screen.Recording.2025-05-22.at.2.29.56.PM.mov

The place to intervene is around here, but we don't yet have an easy way to add something that is morally is_value_position(node) to this condition:

ark/crates/ark/src/lsp/completions/sources/composite.rs

Lines 77 to 78 in 2013ca1

	if completion_context.document_context.node.is_program() ||
	is_identifier_like(completion_context.document_context.node)

(Side note: I also noticie some duplication in the above completion lists 🤔. I have some ideas about that. But it's a separate matter, in any case.)