Adds a pair of leader/follower iterators to ChapelHashtable and a helper method,
and calls them from the Set `these` iterators.

The implementation does frequent recomputation of how to divide up the iteration
space to find full slots.  This can probably be improved on, but I wanted to
take a snapshot of what worked before attempting it.

It also does not help when zippering with arrays yet.  That is also a next step

----
Signed-off-by: Lydia Duncan <[email protected]>

Test comes from issue chapel-lang#15824 and checks zippering two sets of the same length
(but different contents).  Sorts the output so that it is consistent instead of
potentially racy.

----
Signed-off-by: Lydia Duncan <[email protected]>

----
Signed-off-by: Lydia Duncan <[email protected]>

This is a first step towards being able to zip sets with other types instead of
just other sets

----
Signed-off-by: Lydia Duncan <[email protected]>

This meant always computing the number of chunks and which chunk we were
ourselves, since the array iterator expects the other components of the tuple to
be ranges as well (so we can't use followThis to send that information) and the
argument list for the follower is supposed to match the argument list of the
leader aside from followThis.  This is unfortunate, but given the sizes I
checked, seemed to be okay.

----
Signed-off-by: Lydia Duncan <[email protected]>

Covers:
- converting the set to an array and then zippering
- zippering with the array leading
- zippering with an array when the set leads

With the prior commit, all three work (though toArray already worked prior to
this effort).

----
Signed-off-by: Lydia Duncan <[email protected]>

The original is only failing on one of the two machines I checked, so I'm
curious if this also fails there or if it behaves as expected (it passes on the
machine where the original still works)

----
Signed-off-by: Lydia Duncan <[email protected]>

----
Signed-off-by: Lydia Duncan <[email protected]>

I'm not sure if this will actually allow us to check for unequal lengths, but
try it and see

----
Signed-off-by: Lydia Duncan <[email protected]>

This argument indicates the size of the leader in the iteration, so that we can
validate if the iteration space is uneven.  It has a default value so that the
non-zippered case is not as annoying to write

Update various tests to use this iterator instead

----
Signed-off-by: Lydia Duncan <[email protected]>

- document new iterator
- remove commented out writelns
- restore old behavior to Set's these iterators, since we're using a different
  iterator
- comment an if statement in _determineEvenChunks
- close a parenthetical in an older comment in ChapelHashtable

----
Signed-off-by: Lydia Duncan <[email protected]>

It'll present a cleaner interface to users

----
Signed-off-by: Lydia Duncan <[email protected]>

Covers:
- Zippering with a shorter array as leader
- Zippering with a shorter array as follower
- Serial version of the new iterator
- Standalone version of the new iterator

----
Signed-off-by: Lydia Duncan <[email protected]>

----
Signed-off-by: Lydia Duncan <[email protected]>

…f it

Would have added a warning to all the overloads, except for the bug reported in
all the right contexts, whether the bug or the unstable warning gets resolved
first.

----
Signed-off-by: Lydia Duncan <[email protected]>

Use a halt wrapper instead of `__primitive("chpl_error"`, especially because
this is a new error instead of trying to match errors that the compiler can
insert.

While doing that, made the error messages clearer about what was wrong and what
should be done instead (and hopefully head off confusion for the follower case).
And add a test locking in the error message in the leader case

----
Signed-off-by: Lydia Duncan <[email protected]>