Modify some functions to alleviate overflow / precision errors

Modify some functions to alleviate overflow / precision errors

'session_frame_to_track_frame()' and its complement, 'track_frame_to_session_frame()' both accept a double and an ARDOUR::framepost_t (int64_t). For convenience these both get converted to long double. However, the functions are often called with very large values (ARDOUR::max_framepos). When this happens, rounding and precision errors can cause overflow issues. This is an attempt to alleviate that problem. Not sure if it's strictly necessary to accommodate negative values - but I'm assuming these could happen if 'speed' was negative (reverse play?)

This is the reason why 'Select All Objects' (and similar functionality) don't work for some users.