Remove automatic detection for partial compression

[marcospb19]

Fixes #285 and fixes #282 with a feature regression.

What is the problem this feature used to solve?

Ouch infers compression formats from file extensions.

ouch compress file file.tar.gz

From the output file, Ouch infers the formats [Tar, Gz].

---

However, sometimes you want to compress a file that is already compressed, for example:

ouch compress file.tar file.tar.gz

Again, Ouch infers the formats [Tar, Gz] and compresses file.tar with these formats in mind, this is a problem because the end result is an archive of type [Tar, Tar, Gz].

Ouch should, in some way, enable the user to compress this just using [Gz] (I'm calling this partial compression, because the archive is already partially compressed).

The feature

I implemented (with the code removed by this PR) a tiny automatic detection for when formats match.

Problems with automatic detection:

1. Current implementation is flawed (thanks to me, a lot of false-positives and format mismatchs).
2. It's counter-intuitive and unpredictable behavior (especially for first-timers).
3. Users don't notice when it happens because we already log so much.
4. Does not work with stdout and stdin, because those aren't named.
5. Cannot force partial compression without renaming files manually.

Solution

Regress this feature and implement --format flag.