gif: fix incorrect loop_count calculation

[jsm174]

This PR supersedes #725 with new logic.

The GIF89a spec has no concept of looping. When the NETSCAPE extension is present, the stored value is the number of repeats after the first play.

As of now, if the NETSCAPE extension is not present, loop_count will be 0 which means users trying to use this value would think they need to loop indefinitely.

The value for the IMG_PROP_METADATA_LOOP_COUNT_NUMBER property could be:

• No NETSCAPE extension: loop_count = 1 (play once)
• NETSCAPE repeat=0: loop_count = 0 (infinite)
• NETSCAPE repeat=N: loop_count = N+1 (total plays)

This also updates the encoder to skip writing the NETSCAPE extension if loop_count=1, ie:

• loop_count == 1: skip NETSCAPE extension entirely (play once)
• loop_count == 0: write NETSCAPE with repeat=0 (infinite)
• loop_count > 1: write NETSCAPE with repeat=loop_count-1

see here, here, and here

(If this is not acceptable, and we need a new IMG_PROP_METADATA property, I can modify this commit. At the end of the day, I just need an accurate way to know if I need to loop, and if so, how many times.)

[jsm174]

➜  ~ gifsicle --no-loopcount base.gif > 1.gif
➜  ~ grep -aob "NETSCAPE2.0" 1.gif
➜  ~ gifsicle --loopcount=1 base.gif > 2.gif
➜  ~ grep -aob "NETSCAPE2.0" 2.gif
22:NETSCAPE2.0
➜  ~ gifsicle --loopcount=2 base.gif > 3.gif
➜  ~ grep -aob "NETSCAPE2.0" 3.gif
22:NETSCAPE2.0
➜  ~ gifsicle --loopcount=0 base.gif > 0.gif
➜  ~ grep -aob "NETSCAPE2.0" 0.gif
22:NETSCAPE2.0

I guess some of the confusion is that it's not really loop count but "repeats after the first play".

With what I provided, SDL_image would now abstract this concept away.

So I'm definitely fine with changing to IMG_PROP_METADATA_NETSCAPE_EXTENSION, just let me know what the name should be and what type of value it should be.