Update TIFF conversion to support more bit depth.

[mayeut]

Update #322
Update #264

This support 1, 2, 4, 6, 8, 10, 12, 14, 16 bits TIFF both in reading / writing.
Includes (need some testing though) support for BIG ENDIAN (#524)
All tests are OK (well, rounding changed when doing 16 -> 12 bits conversion so need to update non regression data)

[szukw000]

@mayeut ,
my tests on PPC64 (resp. LINUX-64) were OK.

BUT: you have forgotten to apply the patch in #518. I had to apply #518.

The difference between the old code and the patch #518 can be visited by

http://home.arcor.de/szukw000/cio.txt.zip

For tiftoimage() I have made a small change. Instead setting tBps and tPhoto to 0,
I use the variable 'fails' and show the original values for tBps and tPhoto in case
'fails == 1".

winfried