Add WartRemover and fix a reasonable set of warts

[sritchie]

Please feel free to ignore if you don't find this helpful, or we can back out the set of changes that you do like.

This PR adds support for WartRemover, a nice scala linter: https://www.wartremover.org/

My motivation here was to make sure that all public methods and vals had declared return types, as these are very helpful in understanding Rainier, and in checking binary compatibility with mima once/if Rainier starts doing that.

The changes I completed here were:

• adding return types
• removing implicit toString conversions, in favor if the s"" macro style
• adding extends Product with Serializable to sealed traits - this makes the error messages much nicer, as the compiler can actually infer the particular subclass of the sealed trait, vs going all the way back to Product with Serialiable (see here for a nice discussion and rationale: https://nrinaudo.github.io/scala-best-practices/adts/product_with_serializable.html)
• I did have to add ONE trait, EncoderFrom, since once I added a proper return type to Predictor.apply, the Encoder return type prevented anyone from calling .fit on the result, specifically this from RegressionBenchmark:

def build(): RandomVariable[Real] =
    for {
      betas <- RandomVariable.fill(k)(Normal(0, 1).param)
      sigma <- Uniform(0, 10).param
      _ <- Predictor[Double]
        .fromVector(k) { vec =>
          val mean = Real.dot(betas, vec)
          Normal(mean, sigma)
        }
        .fit(data)
    } yield sigma

[codecov-io]

Codecov Report

Merging #393 into develop will decrease coverage by 0.42%. The diff coverage is 37.73%.

@@            Coverage Diff             @@
##           develop    #393      +/-   ##
==========================================
- Coverage    47.32%   46.9%   -0.43%     
==========================================
  Files           88      89       +1     
  Lines         3275    3307      +32     
  Branches       150     151       +1     
==========================================
+ Hits          1550    1551       +1     
- Misses        1725    1756      +31

Impacted Files	Coverage Δ
...main/scala/com/stripe/rainier/compute/Target.scala	91.66% <ø> (ø)	:arrow_up:
...ore/src/main/scala/com/stripe/rainier/ir/Ops.scala	100% <ø> (ø)	:arrow_up:
.../scala/com/stripe/rainier/core/Combinatorics.scala	90.9% <ø> (ø)	:arrow_up:
...main/scala/com/stripe/rainier/core/Predictor.scala	0% <ø> (ø)	:arrow_up:
...re/src/main/scala/com/stripe/rainier/core/Fn.scala	0% <ø> (ø)	:arrow_up:
...ain/scala/com/stripe/rainier/compute/RealOps.scala	78.81% <ø> (ø)	:arrow_up:
...in/scala/com/stripe/rainier/compute/Compiler.scala	100% <ø> (ø)	:arrow_up:
...com/stripe/rainier/bench/RegressionBenchmark.scala	0% <ø> (ø)	:arrow_up:
...ain/scala/com/stripe/rainier/core/Likelihood.scala	37.5% <ø> (ø)	:arrow_up:
.../main/scala/com/stripe/rainier/core/Discrete.scala	84.52% <ø> (-1.02%)	:arrow_down:
... and 32 more

---

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 0521f70...b9727b2. Read the comment docs.

[sritchie]

@avibryant, any thoughts on this one?

_{Sent with GitHawk}

[avibryant]

@sritchie sorry for the delay - I'm definitely bullish, but won't have time to review until monday or tuesday.

[sritchie]

@avi-stripe, just wanted to ping - would you mind giving this a look? I think there's only one real type question, otherwise this should be pretty routine.

Cheers!

[sritchie]

okay, @avi-stripe , this should be good to go.