new approximate inverse code for GMP

[Paul Zimmermann]

       Dear all,

I've revised my note on fast approximate inverse for GMP.
See <http://www.loria.fr/~zimmerma/papers/index.html#invert>.
The code is available too: the mpn_invert2 routine in file
invert.c could replace the mpn_inv routine from 
<http://www.swox.se/gmp/devel/mpn/generic/inv.c>,
except it returns a result which may be 1 bit too small.

It would be great if someone could check the proof of Lemma 2.

Paul

PS: don't look at sections 3 (approximate quotient) and 4 (quotient with
remainder). They are obsolete with the "New truncating division interface and
code" on <http://www.swox.se/gmp/devel/>.