Translate string search-and-replace operator

[clarity20]

Main idea:

${str/x/y}  -->  re.sub(r'x', 'y', str)
${str//x/y}  -->  re.sub(r'x', 'y', str, count=1)

POSIX globs -- N.B., these are not regexes! -- need to be ported to pythonic regexes -- not globs:

abc*  -->  r'abc.*'
ab?c  -->  r'ab.c'
abc[def]  -->  r'abc[def]' (no change)

Extended globs are transformed by moving the operator to the end -- I don't know about !, however:

+(aa|bb)  -->  (aa|bb)+
?(aa|bb)  -->  (aa|bb)?
*(aa|bb)  -->  (aa|bb)*
@(aa|bb)  -->  (aa|bb)@

The real challenge is when expansion is needed. Suppose t='+(xx|yy)' Then the following is NOT correct:

str=ddddxxdddd
echo ${str/$t/qqq}   -->  print(re.sub(t, 'qqq', str, count=1)) 

because to achieve the same effect in python we would need to adjust t: t='(xx|yy)+'.
Probably the best solution is to alter t on the fly when it's needed in calls to re.sub():

re.sub(t, ...)  -->  re.sub(to_regex(t), ...)

for a suitably defined to_regex(). But is that worth the trouble?

[clarity20]

As to the last issue raised above, it should be up to the coder to get t right in a pythonic context. Heaven forbid that t be used for both regex and non-regex purposes; even if it works in bash -- and the transpiler should assume the input code behaves as the developer intends -- we cannot translate it directly to python, as we would need +(...) in some places and (...)+ in others. That's a problem to be solved by the developer, not the transpiler.

[clarity20]

Resolved by #84