Use restore_tester_fmt.c to find and correct restore issues

Things I know about. Add more as they are seen/fixed, since some of these should get done in core also.

• rules. When restarting, the words are restarted for whatever rules were being worked on. (Fixed 2 bugs https://github.com/magnumripper/JohnTheRipper/commit/2110a37528fb936f52ff789a0f5d27f05f09b5d1 and https://github.com/magnumripper/JohnTheRipper/commit/62310e8b5a9aefc2b31da43e167eedcd280fb6a5)
• rexgen does not resume #570 (fixed but can not use until v1.4.0 of rexgen is released).
• rexgen-hybrid only resumes the base-word #570.
• Multi-salt mode restarts all salt work over again (#2087)
• single mode
• markov mode
• mask mode Restore is almost right. Just the very last word which was properly done is started with. (fix found, pending verify)
• wordlist mask hybrid was also 'off by one', i.e. replaying last tested word. (fixed, pending verify)

Some of these could be handled as 'hybrid' type blocks added to the end of a .rec file (I think). Things like the rules and multi-salts redoing work, could be addressed in that manner. The 'normal' V4 processing of the rec file gets us back to the right words, and then the hybrid extra block appended provides additional information on how far to skip ahead inside the subwork done on the word or set of words, so that prior completed work is not replicated.

The restore_tester_fmt is great for determining exactly what was processed, written to .pot files, etc. So that we know 100% that we restart at proper location, not before, and CERTAINLY not later in the word stream.

Start with restore_tester_fmt.c and use it to examine restart behavior for the listed rules, rexgen, multi-salt, single, markov, mask, and wordlist-mask modes. Compare the processed and recorded work across the relevant .rec and .pot files. Done means each mode resumes at the proper location without repeating or skipping word-stream work.