regex_revalidate c++, lifecycle messages, MISS support

[traeak]

A direct port of regex_revalidate to c++ with added targeted messages:

traffic_ctl plugin msg regex_revalidate config_reload
traffic_ctl plugin msg regex_revalidate config_print

The config_reload option allows only regex_revalidate to be reloaded outside of a traffic_ctl config reload
The config_print option will dump current active rules and times to traffic.out (or stderr)

This version uses a std::vector<> of rules which benchmarks faster than the c linked list implementation.
Also an attempt was made to clean up rule loading and pruning.

Based on #7322, the following mode will be available (now WIP):

• Last rule wins, which allows rules to be changed or even unloaded through expiration.
• Any changed rule will have its revalidation activation time reset.

Resolves #7322

Also added the ability to optionally add the keyword "MISS" at the end of a configure line to force an asset refetch instead of just an IMS check.

[traeak]

[approve ci autest]
I think this is legitimately busted for now....nope not busted, behavior changed which allows an existing rule end date to be changed.

[traeak]

I'm not a big fan of all the rvalue boilerplate for the Invalidate struct.
Should I refactor to use tscore/Regex ?
For the time being I created yet another regex.h/regex.cc, additions made to efficiently extract the needed substrings.
Refactoring/extending tscore/Regex.h to replace many of these custom classes should be done.
the config_handler function has the bulk of the logic changes.

[traeak]

I think I will need to add a backwards compatibility mode so that mixed environments can be consistent...not necessary....

[traeak]

Changed requirements based on a compelling end user/limited self service use case from #7322

[bryancall]

[approve ci]

[bryancall]

@ywkaras is going to take a look at this.

[ywkaras]

It seems convoluted to schedule this, then ignore the event it generates, when timed_reload is false.

[traeak]

convoluted is correct. Even with timed_reload disabled any expired rules will still be pruned out so this needs to still be run.

[ywkaras]

Did you do git mv regex_revalidate.c regex_revalidate.cc before you started editing? This might have been a faster review if it was easier to see what was changed.

[traeak]

perhaps. the changes are fairly destructive regardless.

[ywkaras]

Don't the two continuations need the same mutex since they both access pstate?

[traeak]

i believe that wouldn't allow the invalidation and the config continuations to run asynchronously. That's what the atomic locks for pstate does. I'm guessing this might break if global plugin reloading is enabled?