Interrupts in completion, traps in _main_complete

[Bart Schaefer <schaefer@xxxxxxxxxxxxxxxx>]

On Dec 7,  4:21pm, Peter Stephenson wrote:
}
} That's probably the way to go --- my gut feel is the behaviour of the
} other form of trap is too far from what we need to make it a good
} starting point.


I took a look through _main_complete and it does quite a bit of saving and
restoring of global completion settings e.g. in order to implement zstyles.
None of it is nicely isolated, i.e., we're going to need several always-
blocks -- and/or possibly some things declared "local", which I'm not sure
why was never done in the first place -- or else a significant rewrite.