kernel: tweak error messages to say "an object"

[fingolfin]

... instead of "a object", and similar.

This makes some progress on issue #3400 and is otherwise about the least import PR we had in a long time, I guess ;-).

[ChrisJefferson]

I understand the reason someone would want this, but it's going to break the tests of lots of packages, and make it hard to easily write tests with error messages which work in versions of GAP both before and after this patch.

I don't have a good idea on how to fix this. A configuration option feels overkill.

[fingolfin]

I actually do not believe this will break a single package test:

1. These errors concern validation of GAP kernel functions; no package test ought to trigger any of those.
2. To verify this, I grepped through all deposited packages plus a ton of package git repositories I have on my computer, using rg -F '(not a ', and as far as I could tell, none of them trigger any of the affected kernel error messages. There were only a few hits anyway; those in .tst files either were inside comments (for atlasrep), or came from package functions (for datastructures).
3. In GAP 4.11, we completely overhauled these errors (by adding those Require* macros, and also by changing many TNAMs); it's very likely that those changes would have broken packages which this PR here would break. But none were broken, which seems to support my point 2.

[ChrisJefferson]

Yep, I was too cautious.

[coveralls]

Coverage decreased (-0.04%) to 84.439% when pulling b3e0730 on fingolfin:mh/error-article-inflection into 3d9c302 on gap-system:master.