Optimize order of provers to reduce proof time

[yannickmoy]

Use of the tool SPAT showed that a lot of time was spend by CVC4 on some VC
that Z3 could prove easily. Changing the order was not a home run though, as
Z3 runs a long time on some VC in sparknack-utils.adb. Solution is to
special-case this unit to run the provers in the other order. On my laptop,
running time goes from 1 mn 50 s to 1 mn 3 s.

[rod-chapman]

Does the --replay flag work reliably in Community 2020?

[rod-chapman]

Looks good to me. Can we re-enable the --replay flag as well now?

[yannickmoy]

Does the --replay flag work reliably in Community 2020?

you're referring to the crash you got with Community 2019? yes it should be gone.

Can we re-enable the --replay flag as well now?

First, you'll need to commit the Why3 session files, and then people will need to use Community 2020 for it to work. Not sure this is worth doing, given the drawbacks, especially the fact that people who use another version won't understand why it does not work at all.

[yannickmoy]

btw you could have a Makefile with a replay entry, for people who want to use replay. but not impose it in the project file. And this way people can also generate their own session files, and use replay for subsequent runs.