API for unbiased and saturating casting

[mattfel1]

I added nodes and chisel backends for saturating ops by enclosing them in <> (like <+>) and unbiased rounding by appending & (like *&) but I think there can be times when we just want to reduce the precision of a number and have control over whether the cast will saturate or truncate the bits.

For saturation casting, is this the intention of to[T] vs as[T] or are these just supposed to handle sign extension? The backend already exists for whatever kind of conversion, but I think right now I sign extend when upcasting and I think I default to truncation for downcasting.

What should the syntax be for unbiased rounding upon downcasting? I already have chisel for this too but I think we need a node to plug it into.

[dkoeplin]

val x: A 

// Insert an explicit cast from A to B <truncate, roll over> on <reduced precision, under/overflow>
x.to[B]  

// Reinterpret the bits of A as being of type B. Same as casting from (A*) to (B*) in C/C++
x.as[B]

e.g.

val x: Float = 128.5f
val y = x.to[Byte]  // Rolls over, truncates to -128

val z = x.as[Int]    // 0x43008000 => 1124106240

At the moment we don't have a syntax for anything besides truncating/rolling over in type conversion.