gsw_ct_freezing / gsw_t_freezing wrapper functions

[hylandg]

gsw_ct_freezing / gsw_t_freezing are now wrapper functions around the specific gsw_ct_freezing_poly / gsw_t_freezing_poly and gsw_ct_freezing_exact / gsw_t_freezing_exact functions. The wrapper functions have an optional logical flag (exact) that allows selection of the specific _poly (exact=.false. [default]) or the _exact (exact=.true.) functions. This allows maximum flexibility - the specific _poly or _exact routines can be called directly, or in a programatically controlled fashion via the wrapper.

[hylandg]

Made _exact the default in commit d01d63a (and I should have made this comment and the following at that time).

Note that the _exact versions are for internal use only (they are not and will not be listed in the official documentation - for reasons I still don't understand). I created them specifically to make the code more readable - all the Fortran GSW functions call either the _exact or _poly routines. Users, however, will call the _poly version, or gsw_ct_freezing or gsw_t_freezing for the _exact version.

[PaulMBarker]

The poly function is ridiculously accurate, from memory, the difference between the exact version and the polynomial version is less that the errors made in the fitting of the original data for the exact.

The poly version is so much quicker to run.