comctl32_tests: Janitorial: write-strings warning fix

[Andrew Talbot]

Robert Shearman wrote:
> 
> Absolutely. We should avoid regressions at all costs, but making literal
> strings non-writable should help to uncover bugs.
> 

Of course, what I am doing with the structure members is to say "These have
been declared as writeable (LPSTR pszXYZ); I don't know whether they are
meant to be writeable, so I dare not change them to LPCSTR pcszXYZ. In
which case, I am cutting the compiler noise when looking for opportunities
to declare suitable strings const (using the -Wwrite-strings switch). The
question is: is this wise, or am I masking potential bugs?