Fix switch fall-through in crn_decompress_block

[zellski]

Do not immediately overwrite our DXT5-decompressed pixels with a cCRNFmtDXN_XY interpretation.

[illwieckz]

Hi, as one of the maintainer of the @DaemonEngine fork (see https://github.com/DaemonEngine/crunch), I'm interested in merging the fixes people submit if those fixes are not so much intrusive and/or if they fixes obvious bugs that has to be fixed in any way.

@zellski can you confirm

• your change is not introducing behavioral change like making the produced format incompatible with the original one even if the original one would be faulty;
• the fix has to be implemented by upstream one day in anyway because it's a bug that has to be fixed;
• there is a low probability to see upstream implementing the fix another way;
• there is no chance in your own humble opinion that upstream would fix the bug in another way that would introduce a behavioral change with your own fix.

If you can answer Yes to all those questions I may merge your fix in @DaemonEngine fork which is actively maintained to bring fixes while keeping in sync with upstream, hoping for a merge one day, see DaemonEngine#2. Thank you in advance.

[zellski]

Uh, no, I can't confirm all that, but it's a bugfix and you're free to use it.

[illwieckz]

@zellski by bugfix do you mean the current code (without your patch) is always wrong?

[ghost]

I do not understand the code behind, since I don't know the algo and format, but here, a compiler would warn about fallthrough, and it's the only case in which there's a fallthrough.
To me, this is clearly a safe bugfix. That is, I'd rather test in debug situations if any problem arise than let this sleeping.