Refine API

[aleasims]

Changes

• ByteUnit/CoreUnit/TimeUnit now use u64 instead of i64. It makes no sense to store them as signed, especially because parsing negative units always results in error.
• Serialization helpers are now exposed in public API. This is needed because these types are public fields in different models, which are in public API.
• Re-enabled linting and testing in CI. A bunch of tests were fixed. Old e2e tests were ignored. Clippy warnings fixed.
  UPD. moved to separate PR

Notes

Doing all this revealed a number of questions. They are not resolved in this PR, however I will list them here for further discussions.

• Exposing units as public enums makes all of their variants public, which allows to create invalid {Byte,Core,Time}Unit::String variants. Probably private implementation pattern should be used here for better API:

  // public type
  pub struct ByteUnit {
      // private inner field
      inner: ByteUnitImpl,
  }
  // private implementation
  enum ByteUnitImpl { ... }

• Multiplying unit value by factor may overflow and result in a panic. Although it's very unlikely to happened on real input, still probably should be handled.
• There is an inconsistency between bytesize::ByteSize and gevulot_rs::models::ByteUnit. We use "megabyte" to represent mebibytes, but bytesize separates them. Because of that ByteUnit<DefaultFactorOneMegabyte>::from(1) and "1MB".parse::<ByteUnit>() are not the same.
• CoreUnit doesn't implement fractional parse, however it was mentioned in tests. This doesn't work: "1.5cpus".parse::<CoreUnit>(). Do we want this?

[aleasims]

@miax-gevu thanks for the review
@trusch I'd really like you to take a look at this before merging (cause there are some open questions) when you have a bit of free time

[aleasims]

To simplify review I moved all unrelated changes to separate PR:

• Enable testing and linting in CI workflow #54

This PR now contains only serialization helpers changes. (force-pushing to keep it clean)
The questions in PR description are still actual.