Add XBM decoder

[mstoeckl]

This change adds a decoder for XBM files, covering the easy part of #2410 .

My motivation for adding this is that, even though the XBM format is old and inefficient, .xbm files still stick around on Linux systems and I'd like for image viewers using the image crate to be able to interpret them.

Since XBM is not a very common format, the decoder is disabled by default and is not included in the default feature list. (I only see XBM files used for icons and cursors, for which PNG and binary PBM are better choices.)

I've tried to make this implementation as reliable and unlikely to need further improvement as I can:

• Is written in pure Rust, with no unsafe.
• Has no dependencies.
• Does not allocate any memory.
• Does not panic or unwrap.
• Makes a single pass over the input, does not require the reader to implement Seek, and runs in O(n) time.
• Provides error messages on parse failure indicating where the failure occurred.

Note: this PR was in large part made by looking at how the other image formats are handled. I was not able to test the xbm fuzzer due to a random LLVM symbol error that I am still trying to figure out (Edit: updating libfuzzer-sys version to 0.4.9 in fuzz/Cargo.toml fixes it for me. Edit 2: have now fuzzed this for a few hours using a seed corpus of valid images, without crashes).

I plan to make a PR for XPM decoding later, depending on what feedback and improvements this gets. (The XPM decoder has already been written and is at https://github.com/mstoeckl/image/commits/xpm/, but I'm holding off on further improvements until I know what is needed. I could also make a parallel PR for it now, if you'd like.)

---

Update: for draft branches using the possible new image-extras repository and decoding hook framework, see https://github.com/mstoeckl/image/commits/xbm-hook and https://github.com/mstoeckl/image-extras/commits/xbm .