Incorrect/inconsistent reported compression percentage for empty files

[marcusmueller]

Describe the bug

zstd reports that compressing an empty file (i.e. 0 bytes) (which yields a 9 B archive) is a compression ratio of 900%. It's not.

It does, however, correctly, report that compressing a single-byte file to a 10 B archive is a compression ratio of 1000%.

While the compression ratio for empty files is essentially meaningless, the compression ratio 0->9 B is actually worse than 1->10 B, and it should be reported as such.

To Reproduce

$ touch empty
$ zstd empty --no-check -o empty.zstd
empty                :900.00%   (     0 =>      9 bytes, empty.zstd)

Expected behavior

Reporting inf %, or not reporting a compression ratio at all.

Desktop (please complete the following information):

• OS: Linux x86_64, Fedora 31
• Version *** zstd command line interface 64-bits v1.4.4, by Yann Collet ***
• Compiler: Fedora standard, so gcc (GCC) 9.2.1 20190827 (Red Hat 9.2.1-1)
• Flags: Fedora default, -O2 -g

[terrelln]

Yeah, our logic takes compressed size / max(input_size, 1). I'd be happy to accept any PR that fixes this to handle input_size == 0 specially.