<regex>: Some _Matcher3::Skip() cleanups

[muellerj2]

This PR performs some cleanups to _Matcher3::_Skip() that I don't want to mix with the performance work in _Skip() to close the big gap with Boost.Regex for disjunctive regexes.

Each cleanup is split into its own commit.

• _First_arg => _First: I think the local variable was named _First_arg to avoid shadowing the member _First of _Matcher3. But this member was removed a while ago, so we can rename the local variable to the more common name _First now.
• add const: The lack of const is pervasive in <regex>. I have to admit I didn't really cared about this until now because there were more much important problems to solve, but we should change this especially for any pointers to NFA nodes in the matcher. This adds const to _Skip() and applies the minimally necessary const propagation to _Do_class().
• correct _Skip signature: _Skip() had the wrong signature: Its callers pass iterators of type _It and all functions called internally return _It or iterators of the same type they are called with. But even so, the function signature stated that _Skip() would take iterators of type _BidIt and return an iterator of type _BidIt.¹
• name all parameters in declaration: I'm not sure whether we should assign names to the parameters in function declarations or not. But assigning parameters to some of them but not all is strange.

Footnotes

1. I think _BidIt is intended to refer to the wrapped and _It to the unwrapped iterator type. But this intention hasn't been followed through on. This signature change fixes one inconsistency, but there is at least another one when assigning the capturing group iterators to match_results in _Matcher3::_Match(). And the apparent consequence of these inconsistencies is that the regex_match()/search()/replace() wrappers always set _BidIt and _It to denote the same type. ↩