Support character classes in glob authz rules

Proposal
Change the way section names are parsed in authz files, so that anything between the initial [ and the last ] are treated as the section name/rule pattern. This will restrict the use of inline comments on rules, but will allow fuller use of glob patterns.

Discussion
Wildcard rules in Subversion's authz file use apr_fnmatch() to match patterns to path segments. APR's matcher supports character classes ([A-Z] etc.), however, there is no way to use them in the authz file. For example, given this rule:

[:glob:/**/*.[Dd]oc]
* = rw

the pattern will be parsed as "/**/*.[Dd", which is probably not what the user intended and matches, for example, "/path/x.[Dd" instead of the expected "/path/x.doc".

This limitation was present in the authz rules since their inception and is a side effect of how Subversion's config parser parses section names. See: parse_section_name() in libsvn_subr/config_file.c.