Tohil tclobj integer math always at 64 bits; Faster Python-to-Tcl numeric conversions; TclProcs now return tclobj by default

[lehenbauer]

• Perform all integer operations using Tohil tclobj objects using 64-bit integers, even on 32-bit machines. (It was already 64-bits on 64-bit machines.)
• When creating a new Tcl numeric object from a numeric Python object, we now do so much more efficiently (see below).
• Make TclProcs default to returning tclobj data type instead of string. (Fixes issue TclProcs default to returning str but call defaults to tclobj #72)
• Expanded number ranges for hypothesis framework testing of tclobj math.

When creating a new Tcl object from a numeric Python object, previously tohil always converted it to a string, which thereby supported both Tcl and Python's arbitrary precision arithmetic, but would be slower than using access functions that worked on machine-native numeric types when those types were large enough to handle the numbers being represented.

Tohil, now, if the Python object is an integer, tries PyLong_AsLongAndOverflow to obtain it as a C long, and if it works without overflow, uses Tcl_NewLongObj to create the Tcl equivalent. If it overflowed, indicating the number is too large for a long, Tohil tries PyLong_AsLongLongAndOverflow, and if that doesn't overflow, tohil creates and returns a new wide Tcl integer through Tcl_NewWideIntObj.

Likewise if the Python object being converted is floating point, it is brought into Tcl using PyFloat_AsDouble and Tcl_NewDoubleObj.

Now, only if the number is wider than a long long, or complex, is it converted to a string for use by Tcl.