fix mishandling of ^ inside an expression

[marler8997]

Currently, if a ^ character appears inside a regex pattern and is neither the first character of the expression nor the first character in a character class, then it will be "compiled" to a BEGIN object, but will be interpreted by the pattern matching code as a CHAR object (in the matchone function). This means it will try to match the text against the BEGIN objects u.ch field which will be uninitialized memory.

For the fix I checked what the Python regex implementation did in this case, it appears that it does not match the ^ character inside an expression, here's some python code that demonstrates this:

import re
r = re.compile("a^")
for i in range(0, 255):
    text = "a" + chr(i)
    print("checking '{}'".format(text))
    assert(not r.match(text))

Along with the fix I also added a regression test that ensures that a^ never matches any sequence of two characters starting with a.

[marler8997]

ping @kokke, if you have time I'd like to get some of the trivial PRs reviewed. This one is a single-line bug fix to the implementation along with some additional testing. Another trivial bug fix is #64

If you review/merge these 2 PRs, I can then maybe try to identify the other bugs that are preventing the random tests from working so we can enforce that they pass on the CI (see #66). Getting the tests working again is the first thing I'd like to help with since it's the sure fire way to enforce that any other changes are correct.