[BCSharp]

PythonAst never got a location, so the code object of module, exec and eval code spanned line 1 only: co_firstlineno was always 1, and setting f_lineno from a trace function failed for any line but 1.

This PR sets the span of PythonAst when parsing finishes, from its first statement to the end of its body, which gives co_firstlineno the same as in CPython 3.4.
It also places the entry sequence point at that start, so the 'call' trace event agrees with co_firstlineno.

I noticed also that IronPython does not handle well modules with the same name, which is quite likely when exec-ing a string (often given a name "<string>"). This is not addressed here.