Change `ta_pointer_start` and `ta_pointer_end` reporting

[kbseah]

These are presently reported for both junction and segment IES features.

However for junction features, which are zero-length, ta_pointer_start equals start coordinate if the IES is bounded by a TA repeat. This results in coordinate for ta_pointer_start being correct for python zero-based numbering system but not for GFF.

For junction features, ta_pointer_end equals ta_pointer_start, which doesn't make sense either.

Maybe just have a ta_offset attribute that states how many bases the start/end coordinates need to be adjusted. This would then be robust against any change in coordinate systems and also be suitable for both insertions and deletions.