API: Add a CacheF() function to control completion cache refresh

[maxlandon]

No description provided.

[rsteube]

Need to give this some thinking when i got the time. I feel like this should already be doable with the timeout (e.g. zero duration) and key arguments.
If anything i'd probably opt for a custom error returned by the key function 🤔 .

[maxlandon]

You might very well be right, and its probable that I've come to open this PR just because I over complicated my own use of the cache in some instances.

So feel free to close if in a few days you still have no clues on this, as it might just be me trying to solve a non-existing problem.

[maxlandon]

Two other things:
1 - I saw you started leveraging lexers for command-line consumption, great idea, curious to see what wonders you'll do with this.
2 - I tried to attack this infamous cobra global comp map, here: spf13/cobra#2012 (comment), if you have anything you'd like to contribute in order to pull this chard out of everyone's foot, that will be my joy !

[rsteube]

Yes, zero timeout duration seems to do the trick.
Writing a lexer is actually something i wanted to avoid, but it is needed for Split.