RFC: Sketch of vector-valued GP functionality

[willtebbutt]

Summary

What: Proposal for a vector-valued GP.

Why: It might be a convenient thing to have around for some use-cases in which people really do think about their multi-output GPs as being vector-valued.

Why: it's easy to provide, because we already have the functionality, we just need to package it nicely.

Proposed changes

Add a vector-valued GP. Not proposing to export at this point in time, because I don't know what I even think about it.

Contains a sketch of the implementation in which the intent is hopefully clear. Does not contain proper testing, and I don't believe that the API is complete.

What alternatives have you considered?

I considered not bothering with this, and I still think that's a good option -- I really don't know whether or not I like this, but I can imagine that it would ease cognitive burden in some common workflows.

Breaking changes

No breaking changes.

[willtebbutt]

Added posterior implementation.

[codecov-commenter]

Codecov Report

Merging #218 (fa1ceaa) into master (df23fdf) will decrease coverage by 1.59%. The diff coverage is 81.08%.

@@            Coverage Diff             @@
##           master     #218      +/-   ##
==========================================
- Coverage   98.02%   96.43%   -1.60%     
==========================================
  Files          10       11       +1     
  Lines         355      393      +38     
==========================================
+ Hits          348      379      +31     
- Misses          7       14       +7     

Impacted Files	Coverage Δ
src/vector_valued_gp.jl	81.08% <81.08%> (ø)
src/finite_gp_projection.jl	100.00% <0.00%> (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 df23fdf...fa1ceaa. Read the comment docs.