[PATCH 2 of 3 v2] tests: use histedit helpers

[timeless]

Augie Fackler wrote:
> Hm. Your heart is in the right place,

I did it because there's another patch where I'm rewriting most of
these hashes...

> but I'm not crazy about the specific implementation.

> We've talked about /having/ limited undocumented revset support
> in histedit - would you be interested in doing that work

I like the general idea, but don't have the focus for it now.

> and then moving most (but not all, as we still need some tests
> that use hashes explicitly)

I left a couple of tests with explicit hashes (very very very few)

> of the tests to that without the helpers?

At some future point...
I'd have to think about it a bit more. One problem is that the
"primary" ui for histedit is a text editor, where it doesn't really
make sense to use a revset (revsets can be surprising, as I learned
when I tried to use them here).

> As an alternative, I'd be interested in some sort of plan filter
> helper, that did something like:
>
> while read line ; do
>   action=`cut -d' ' -f 1 $line`
>   revset=`cut -d' ' -f 2 $line`
>   revsision=`hg log -r "$revset" --template '{node}'`
>   echo $action $revision
> done
>
> That is, write a filter that does revset->hash transformation on a
> flat file, rather than having this stateful-shell-function-helper
> shenanigans. What do you think?

This seems reasonable. Is it ok for it to be shell?