bug#16214: Consistency in dired-, occur-, and grep-mode

[Drew Adams]

While there might be room for some minor alignment, in general
it is not good to privilege standardization too highly here, IMO.

There are superficial similarities, and maybe some that are more than
superficial.  But there are also different purposes and use patterns.

These are quite different modes when you look closely and take all
of what each does into account - its raison d'etre: what it is for.
Each should be handled case by case, with an eye to all of its
features and its overall set of use cases.

Dired, in particular, is extremely rich.  Let us not start hobbling
it in the name of standardization.

With that caveat expressed, I have no big objection to what has
been proposed here so far.  (I would prefer that SPC in Dired
remain what it is, but that's about it.)

But I would strongly recommend that we not go overboard with such
an approach - that would be quite misguided IMO.

The main guide for us should be the full set of features - and how
they interact - for each individual mode viewed on its own.

Let's keep in mind the most important rule regarding consistency:

Consistency *within* a given system/application/realm/area/function
is very important.  Consistency *across* different areas is not so
important - essentially only a nice-to-have, permitted when other
things are in fact equal.

Consistency within an area involves/includes also how its parts
fit together.  Individual parts (e.g. key sequences) should not
be considered only on their own - they are parts of a whole/system.

And even in the latter case, when we allow ourselves some added
consistency across areas, we should always keep that *tentative*.
A better idea that comes up later, based on reasons relevant to a
given domain itself, should trump any such cross-domain tentative
harmony we allowed before that better idea.

IOW, internal cohesion/meaning/relevance is in general a much more
important consideration than is external coupling/consistency/convention.