pj_phi2(): speed-up computation (and thus inverse ellipsoidal Mercator and LCC)

This does not change the numeric values returned by the method,
as far as I could see on a few samplings.

The tricks used save a call to sin() and atan() at each iteration.

This directly affects speed of inverse Mercator and LCC (among others),
in their ellipsoidal formulation.

Timings on inverse Mercator show a 31% speed-up at mid-latitudes
where pj_phi2() needs 5 iterations, and 24% at latitudes close to 0 or
90deg where it needs one iteration.

0 files changed, 0 insertions, 0 deletions