Consider removing act-state

[MattWindsor91]

I've been musing about the possibility of deleting act-state (which mainly only exists to deal with test amplification now). Two options are on the table as to what happens to act-state dnf:

• move to act-c (as it's an operation on C/Litmus postconditions);
• move to c4t (as it's an operation on c4t observation records);
• move to a Python or Bash script in c4-scripts;
• move to a new Go tool, and expose c4t's API publicly.

My gut feeling is to move to c4t, as this means we only have one codebase that has to parse those observation records; this does mean, however, that c4t has to learn a tiny amount about the Litmus language. The fourth option can always be considered if c4t becomes too bulky (I'm beginning to think it's slowly turning into a monorepo, myself).

[MattWindsor91]

Removed (dnf has gone to c4t-obs; the other functionality will be rewritten in c4t as and when necessary).