fix the expected/actual value ordering in a few unit test asserts

[Numpsy]

rather trivial, but still

I certify that I own, and have sufficient rights to contribute, all source code and related material intended to be compiled or integrated with the source code for the SharpZipLib open source product (the "Contribution"). My Contribution is licensed under the MIT License.

[codecov]

Codecov Report

Merging #627 (e723cce) into master (34879be) will decrease coverage by 0.11%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #627      +/-   ##
==========================================
- Coverage   73.35%   73.23%   -0.12%     
==========================================
  Files          68       68              
  Lines        8721     8721              
==========================================
- Hits         6397     6387      -10     
- Misses       2324     2334      +10     

Impacted Files	Coverage Δ
...ICSharpCode.SharpZipLib/BZip2/BZip2OutputStream.cs	78.40% <0.00%> (-1.13%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 34879be...e723cce. Read the comment docs.

[Numpsy]

Not sure what's going on with the coverage results - this shouldn't effect coverage at all, even less so in BZip2OutputStream

[piksel]

No, there is something really odd going on. I have tried fixing it by reducing the amount of test coverage results being uploaded, but it's still bonkers.