Add regex syntax to tokenizer

[squell]

Right now to only give a more precise error message, CC #1350

[bjorn3]

The regex is still passed through glob::Pattern::new() which can error on valid regexes like ^a**$ I think.

[squell]

The regex is still passed through glob::Pattern::new() which can error on valid regexes like ^a**$ I think.

I only considered the case where the program arguments are a regular expression. Since the example given on the sudo blogpost only deals with that (and only the wildcards-in-arguments feature is problematic), I had missed that in sudoers the command can also be a regular expression (🙈); to quote man sudoers:

In sudoers, regular expressions must start with a ‘^’ character and end
with a ‘$’. This makes it explicit what is, or is not, a regular
expression. **Either the path name, the command line arguments or both
may be regular expressions. ** Because the path name and arguments are
matched separately, it is even possible to use wildcards for the path
name and regular expressions for the arguments. It is not possible to
use a single regular expression to match both the command and its
arguments. Regular expressions in sudoers are limited to 1024
characters.

I'll add a seperate test-case and patch for that.

Note: the current code already requires that a command always starts with /, so a command being specified as a regular expression already gives an error:

/etc/sudoers:42:12: expected path to binary (or sudoedit)
squell ALL=^/bin/ls$ -la
           ^

But I've changed that to also complain about regular expressions.

Note that in ogsudo, the following is not a correct way (per the quoted manpage above, and you can also verify this) to write a regex match for command-and-arguments:

squell ALL=^/bin/ls -la$

But it wil not give any error message to that effect (it will try to match a binary with the name ls -la, which you can run with whatever argument you like)