[AutoDiff] Initial Forward Mode AD Support

[bartchr808]

This PR extends on the previous work here on the JVPEmitter (#25954) to make it be able to generate SIL code for the differential function, as well as more correctly implementing the body of the JVP.'

generates a valid JVP and differential. I have yet to add support for control flow and generics. This PR also renames PullbackInfo to LinearMap info since it has a lot of tangent mapping helpers that JVPEmitter can use when creating the differential.

This is still early, so the JVP/Differential emission only runs with the flag -Xllvm -run-jvp-generation when running Swift. With it, you can differentiate basic Float functions with + - / * with no vars like:

@differentiable
func foo(x: Float, y: Float) -> Float {
  return x * x * x * y
}

@differentiable
func binary(x: Float, y: Float) -> Float {
  let a = y / x
  return x * y + a
}

[dan-zheng]

Could you please resolve merge conflicts and rebase into a single commit?
EDIT: "Squash and merge" instead of rebasing is fine as well.

[bartchr808]

@swift-ci please test tensorflow

[bartchr808]

@swift-ci please test tensorflow

[bartchr808]

For time going to run tests now to get this merged in as soon as possible assuming there are only NFC changes that need to be made

[bartchr808]

@swift-ci please test tensorflow

[bartchr808]

@swift-ci please test tensorflow

[bartchr808]

@swift-ci please test tensorflow

[bartchr808]

@swift-ci please test tensorflow

[rxwei]

This is not testing differentiation through protocols at all because you calling a method with concrete Self type Linear.

To differentiate through a protocol requirement, define a generic function that takes an instance bound to a generic parameter that conforms to the protocol, and differentiate that.

func genericFoo<T: DiffReq>(_ t: T, _ x: Float) -> Float {
    t.foo(x)
}
...
valueWithDifferential(at: 5) { x in genericFoo(inst, x) }

[bartchr808]

discussed offline, will address this in the next PR which will support everything else

[bartchr808]

@swift-ci please test tensorflow

[bartchr808]

@swift-ci please test tensorflow

[bartchr808]

@swift-ci please test tensorflow

[rxwei]

Congrats on going through all the hurdles to finish this important step!