V2: refine bend-loss model: transition loss, WGM, bend-induced birefringence

[ameoli]

Follow-up to #36 which added the basic macrobending Marcuse-style A*exp(-R/R_c) model. Real fibers under bending show more physics that is missing:

• Transition loss at the ends of a bend section (mode-mismatch between straight and curved sections, usually modelled with an overlap integral between the straight-fiber and curved-fiber modes)
• Whispering gallery modes: above a critical coupling the light couples back into a cladding WGM and leaks out distributed along the bend rather than in a single macrobend-loss hump
• Bend-induced birefringence: differential stress between inner and outer side of the bend splits the two linear polarisations -- shows up as PDL and a polarisation mode group delay
• Temperature dependence of the coupling (the coating matters above ~120 C)

Primarily matters for tight bends (R < a few mm) and polarimetric OFDR. Not urgent.