Implement index mutation

[jonjohnsonjr]

Add an AppendManifests function that adds entries to the manifests clause of an index.

Fixes #474

[codecov-io]

Codecov Report

Merging #520 into master will increase coverage by 0.08%.
The diff coverage is 70.68%.

@@            Coverage Diff             @@
##           master     #520      +/-   ##
==========================================
+ Coverage    73.1%   73.19%   +0.08%     
==========================================
  Files          92       94       +2     
  Lines        3938     4006      +68     
==========================================
+ Hits         2879     2932      +53     
- Misses        696      703       +7     
- Partials      363      371       +8

Impacted Files	Coverage Δ
pkg/v1/mutate/mutate.go	73.8% <100%> (+3.42%)	⬆️
pkg/v1/mutate/image.go	67.47% <67.47%> (ø)
pkg/v1/mutate/index.go	75% <75%> (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 0392546...d056be0. Read the comment docs.

[jonjohnsonjr]

An alternative to having two methods is something like:

type Appendable interface {
	RawManifest() ([]byte, error)
	MediaType() (types.MediaType, error)
}

type IndexAddendum struct {
	Add v1.ImageIndex
	v1.Descriptor
}

func AppendManifest(idx v1.ImageIndex, adds ...IndexAddenedum) v1.ImageIndex {
  switch v := i.(type) {
    case v1.ImageIndex:
      // add to indexMap
    case v1.Image:
      // add to imageMap
    case v1.Layer:
      // TODO: how can we do this, actually?
    default:
      // log warning for unexpected addendum
  }
  // add descriptor to manifest
}

We could also do type checking against something that just implements MediaType? It's hard to come up with something reasonable here if an index can reference arbitrary objects.

I suppose we could also append anything that implements:

type Appendable interface {
	Size() (int64, error)
	MediaType() (types.MediaType, error)
        Digest() (v1.Hash, error) 
}

Neither v1.Image nor v1.ImageIndex implement Size, but it would be trivial to add. We could also add a method that returns a v1.Descriptor, which would allow objects to maintain annotations if they needed. This would be useful for e.g. a foreign v1.Layer that knows what URLs it should have.

[imjasonh]

Do we want to return here?

[jonjohnsonjr]

We need to add a Blob method to v1.ImageIndex if I want to fix that. I can send a PR for that as well, but as it is this will break our tests if I return an error. I'm pretty sure I do want to add Blob to that interface, but I'd like someone to tell me it's a good idea first :)