Python-gmp

-------------------

* minor cleanups, ensure support for Python 2.3
* fixed misdiagnosis of some argument counts in macro
* SELF_ONE_ARG_CONVERTED (tx to Paul Rubin!)

-------------------

* change ValueError to OverflowError for 'too-large' errors
* fix bug in mpq_pow (negative base, exp. with odd denominator)
(fix now corrected -- _even_ denominator is the error!)
* fixed gcc warnings reported by K. Briggs
* support GMP 4 (but added no GMP4-only functionality yet)
* updated tests to 0.9, better coverage

-------------------

(again, requests & suggestions by great Pearu!)

* raise test coverage 72.5% -> 90.0%
* introduced callbacks (not documented/tested for now;
Pearu will test/support/document in PySymbolic)
* some errors went undiagnosed, caused crash: now fixed
* workaround for GMP bug(?s?) in mpz_fits\_... (?)
* added exposure of mpf\_ sqrt and pow_ui

-------------------

Good feedback from Keith Briggs, some advice from Tim Peters and Fred Lundh ---
thanks all!

* fixed bug of '"%d" where "%ld" was meant' in many places
and other sundry minor warnings given by gcc
* fixed hash (delegating to Python) so mp[nqz](x) will
produce the same value as hash(x) for any Python number x
* workaround for GMP 3.1.1 bug, mpz_root wrongly returning
'exact' for non-exact root if dest==source, which stopped
needed value-error for inexact mpq**mpq operations
* determined correct 'actual precision' of floats
* explicitly stored precision with binary-form mpf's
* extended explicit-bits request to all ->mpf operations
(good in itself, plus, preparing for future MPFR)
* removed the limitation of no binary-form for <0 mpz
* introduced macros to parse args, for conciseness

-------------------

(lots of good ideas from Pearu once more!-)

* fixed silly bugs in kronecker and mpq_abs
* gmpy-level workaround for scan0/scan1 bugs (?) in gmp 3.1.1
* added qdiv; anynum->mpq substituted for all such conversions
(also anynum->mpz and anynum->mpf by analogy, with care!)
* added global.fcoform for optional use of intermediate string in
float2mpf (used for any float->mpf conversion)
* added set_fcoform function for global.fcoform access
* general cleanup of sources; added alloca for msvc++;
- many sundry minor bugfixes & uniformization;
- a little useful refactoring (more would be good...)
* added caching of mpq objects
* power for mpq
* stern-brocot algorithm for mpf->mpq (also exposed as f2q)
- also used for float->mpq
- with stricter tracking of mpf's requested-precision
- added getrprec method to mpf, getrprec module-function
* exposed ceil, floor and trunc methods/functions for mpf's
* changed a couple exceptions from value to zerodivision
* added 'qual' and 'floa' options to gmpy.rand

-------------------

* added jacobi, legendre, kronecker
* added random-number generation, seed set/save, shuffling
* added mpq (at last!-)