epoch::from_duration slight improvements

[gwbres]

Hello @ChristopherRabotin,

this is an inquiry that might get pruned.
i'm trying to see if there are any possibilities to simplify some conversions and construction ops.

Right now the only thing I achieved is a slightly tighter Epoch::from_duration and use it everywhere.

Inquiring #237 and possible traits will most certainly have more impact

[codecov]

Codecov Report

Patch coverage: 100.00% and project coverage change: -0.09 ⚠️

Comparison is base (4d0f6f0) 81.09% compared to head (c8648ef) 81.01%.

Additional details and impacted files

@@              Coverage Diff              @@
##           4.0.0-dev     #238      +/-   ##
=============================================
- Coverage      81.09%   81.01%   -0.09%     
=============================================
  Files             16       16              
  Lines           3783     3760      -23     
=============================================
- Hits            3068     3046      -22     
+ Misses           715      714       -1     

Impacted Files	Coverage Δ
src/epoch.rs	88.94% <100.00%> (-0.11%)	⬇️
src/timescale.rs	97.43% <100.00%> (ø)

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[ChristopherRabotin]

If #237 architecture is adopted, this can simply become:

pub const fn from_duration(duration: Duration, ts: TimeScale) -> Self {
    Self {
        duration,
        time_scale: ts
    }

Then the conversion to/from different time scales only happens when another time scale is requested. For example, if someone did this:

let e = (366.days() + 1.hours()).ephemeris_time() // Create an ephemeris time epoch that's 2001-01-01T01:00:00 ET (because the ET reference is 01 Jan 2000)
let e_utc = e.into_time_scale(TimeScale::UTC) // This is when the computation between time scales happen

Note that I think into_... works here given the Rust naming convention.

As I type this, two other things come to mind:

1. I think it would make sense for Week to be a Unit so one could do 5.weeks() + 6.hours(): a week is always exactly 7 days, and I think we didn't add Week to Unit because it would have been a breaking change
2. The reference epoch should be defined for all TimeScales. The issue still comes with modifiers since there may be some use cases of 365.days().mjd_ephemeris_time() where the reference epoch is Modified Julian Days... but that adds complexity and rigidity because people can't easily define new reference epochs... Maybe something 365.days().ephemeris_time_from(REF_EPOCH) where ref_epoch is also an epoch that's added to the ephemeris_time reference epoch ... I don't know what's best!

[gwbres]

moved to dev-gh-237 since i'm inquiring this topic specifically