bug#50470: 27.1; 'company-mode' 'eshell'

[Jim Porter]

On 3/19/2023 6:34 PM, Stefan Monnier via Bug reports for GNU Emacs, the 
Swiss army knife of text editors wrote:
> I definitely don't want to force preloading that module.
> But maybe that var could have a meaning that's independent
> from completion, thus justifying to move it out of the completion
> extension module?

Well, luckily(?) it turns out my patch wasn't quite right anyway, so I 
completely rewrote it. (In particular, it didn't correctly generate a 
top-level stub if there was a subcommand nested somewhere *inside* an 
argument.)

With this change, we now have a more-general way of preventing commands 
that can cause side effects: 'eshell-allow-commands'. We can let-bind 
that to nil, and then any commands within an argument will signal an error.

Then we just need to disable globbing via a different method (using the 
patch I originally posted), and all is well for this bug.

I also added a couple preliminary patches to fix some semi-related 
issues I discovered while working on this. These could probably go in a 
separate bug, but I'm lazy. ;) The real meat of this change is patch 0003.

0001-Fix-an-edge-case-in-how-eshell-do-eval-handles-let-b.patch
Description: Text document

0002-Simplify-parsing-subcommands-slightly.patch
Description: Text document

0003-Avoid-parsing-some-Eshell-forms-when-performing-comp.patch
Description: Text document