[m-dev.] diff: changes for function mode inference

[Andrew Bromage]

G'day.

Fergus Henderson wrote:

> Andrew, can you please review this change?

Apart from all the merge conflicts you've just caused me, it looks fine.

Just one query: do you have any test cases for this part, or did the
problem only occur when you changed the other stuff?

> modes.m:
[...] 
> 	This change is required
> 	because without it it was incorrectly computing the instmap
> 	delta for the main goal, if the after-goals list was non-empty.

Cheers,
Andrew Bromage