combine WCS and projection type

[xzackli]

This very large PR removes the old projection types, which were redundant with the WCS. Instead, this PR introduces a custom immutable WCS struct,

struct CarClenshawCurtis{T} <: AbstractWCSTransform
    cdelt::Tuple{T,T}
    crpix::Tuple{T,T}
    crval::Tuple{T,T}
    unit::T  # conversion factor to radians. we follow convention and stay in degrees
end

One can choose between this or the C-based WCS.WCSTransform, and using this WCS makes certain parts of the package substantially faster. This struct stays on the stack, consists of 7 numbers, and is definitely more cache-friendly than the kb-sized WCS header we were dragging around. This also means that manipulation of small Enmaps, particularly ones that contain views, isn't dominated by repeatedly copying the WCS header. Although a lot of code changed, I don't think this is breaking.

We are losing some info since the WCS header can have lots of things in it. I should try to deal with this appropriately in the IO routines, probably by providing an option to convert back or follow a WCS template. @guanyilun just a quick clarification, is the default behavior to always convert to POLCCONV to COSMO, and then write to disk as COSMO?

[codecov-commenter]

Codecov Report

Merging #23 (3a9a4cf) into main (85efcb2) will decrease coverage by 3.06%.
The diff coverage is 70.17%.

@@            Coverage Diff             @@
##             main      #23      +/-   ##
==========================================
- Coverage   66.08%   63.01%   -3.07%     
==========================================
  Files           3        3              
  Lines         342      365      +23     
==========================================
+ Hits          226      230       +4     
- Misses        116      135      +19     

Impacted Files	Coverage Δ
src/enmap_ops.jl	59.30% <60.00%> (-2.29%)	⬇️
src/enmap.jl	59.11% <62.50%> (-5.55%)	⬇️
src/enmap_geom.jl	100.00% <100.00%> (ø)

---

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 85efcb2...3a9a4cf. Read the comment docs.

[guanyilun]

Yes I think that's right.