Nice-to-have fix for autogen.sh

[Pete Batard]

Some timings, so that you can feel the pain of Windows users.

Exact same machine (triple boot, with Win7 x64, Slackware Linux 13.37 
and <cough>intosh Snow Leopard). Core2 6400  @2.13GHz with 4 GB RAM.

All done using current libusb-pbatard, starting with a clean tree.
----------------------------------------------------------------------
[Linux 3.0.1 SMP x86_64]

time ./autogen.sh
real    0m12.813s
user    0m5.524s
sys     0m1.404s

time ./configure
real    0m4.304s
user    0m1.489s
sys     0m1.024s


[OS X Snow Leopard, with X-Code]

real	0m18.355s
user	0m9.166s
sys	0m5.011s

real	0m6.852s
user	0m2.355s
sys	0m3.713s


[MinGW32 + msys 1.0 (Win7 x64)]

time ./autogen.sh
real    1m51.493s
user    0m17.783s
sys     0m48.862s

time ./configure
real    1m14.132s
user    0m10.247s
sys     0m34.180s


[MinGW-w64 + msys 1.0 (Win7 x64)]

time ./autogen.sh
real    1m47.656s
user    0m17.783s
sys     0m47.065s

time ./configure
real    1m16.518s
user    0m10.560s
sys     0m33.927s


[cygwin (Win7 x64)]

time ./autogen.sh
real    3m10.458s
user    0m20.536s
sys     0m54.453s

time ./configure
real    2m1.571s
user    0m10.853s
sys     0m37.734s
----------------------------------------------------------------------

And don't get me started on the configure from libcdio, that takes about 
double the time to run, so libusb is actually not that complex.

Regards,

/Pete