Allow running passes multiple times.

[aDotInTheVoid]

Previously, limit was always nonzero, which meant that !limit-- was always false, so the inner loop would never iterate.

Closes #61
Closes #66

alona@Ashtabula:~/dev$ cat issue61.ebnf 
binding-names = {name, ","}, name, ",", name;
alona@Ashtabula:~/dev$ cat issue66.ebnf 
binding-names = {name, ","}, name, ",", name, [","];
alona@Ashtabula:~/dev$ ./kgt/build/bin/kgt -l iso-ebnf -e rrutf8 < ./issue61.ebnf 
Segmentation fault
alona@Ashtabula:~/dev$ ./kgt/build/bin/kgt -l iso-ebnf -e rrutf8 < ./issue66.ebnf 
binding-names:
                      ╭────>────╮
                      │         │
    │├──╭── name ──╮──╯── "," ──╰──┤│
        │          │
        ╰── "," ───╯

alona@Ashtabula:~/dev$ ./kgt-2/build/bin/kgt -l iso-ebnf -e rrutf8 < ./issue61.ebnf 
binding-names:
    │├──╭── name ── "," ──╮── name ──┤│
        │                 │
        ╰────────<────────╯

alona@Ashtabula:~/dev$ ./kgt-2/build/bin/kgt -l iso-ebnf -e rrutf8 < ./issue66.ebnf 
binding-names:
                                     ╭────>────╮
                                     │         │
    │├──╭── name ── "," ──╮── name ──╯── "," ──╰──┤│
        │                 │
        ╰────────<────────╯

I'm not sure how to test this.

[edk0]

Just on a moral level running the pretty steps to convergence shouldn't be necessary to get a correct tree. It's been too long since I worked on this code, but I really feel like that's something we should fix.

[katef]

I agree with @edk0 here, the iteration shouldn't be neccessary. But the loop is there, so let's have it working until it gets removed.

[katef]

Thanks for the PR!