handle exponents greater than 2 digits

if precision is specified, 0 flag is ignored for diouXx conversions
also, print out padding characters for special case 0 value, 0 precision

remove references to long double; it's just not available

write 0's for unreasonable precision requests after number conversion
combine sections of code to print out buffer

fix rest of Chris' bugs, mostly sign and padding placement

if zero value, and precision of zero, nothing gets output

always pad with spaces on the right

Chris' fix for alternate hex and octal mode

rip out statics; go back to fixed buffer size, malloc is just too
painful to get right

don't reinit buffer each time

Chris says that ANSI now says if you specify the precision of a string
you should only look for a NUL in the first "precision" characters.

Chris argues convincingly for a hex representation for 'p'

handle infinite floating point precision

Chris's changes, plus some minor fixes

make `"%4l3d"', `"%*3d", 4', and `%3*d' do the right (wrong) thing
at least it's consistent

fix string precision check

redo putc macro for part of %s, and parsing format string for speed

minor whacks; checkpoint before speed attempt

minor cleanups, move format checking into _cvt

passes basic checks with rewritten floating point

checkpoint; gonna try and rewrite floating point

first version to pass basic tests; rewrite complete

version that attempts to do formatting of floats itsellf

add 'e' format

first cut at ANSI version; also added 'f' format