Reworking attributes to match Ctx

[UnsignedByte]

This reworks attributes to use a system like the Ctx trait for attribute types. This would allow us to use a single get call with any type of attribute. This system currently has a few issues:

• The Attr struct is currently used to parse attributes, but is pretty useless otherwise.
• Ideally, we would like to be able to give different sets of attributes for different attribute contexts (ports vs components).

[UnsignedByte]

@rachitnigam Currently attribute types are specified in the utils directory but that doesn't seem right. I could also move them to ast instead if that seems more reasonable but they are also used in IR...

[rachitnigam]

I don't have strong opinions on where they should live! The utils separation is probably because they are included by both ast and ir but we don't want to include ir in ast.

[UnsignedByte]

Yup, definitely agree with not including ir in AST. My main worry is that although the attribute traits and stuff could go into utils, the actual attributes themselves are somewhat tied to Filament and my impression of utils was that they should be "language-agnostic" compiler utilities... Maybe I will move attribs to AST.

[rachitnigam]

No, the utils are utilities for the Filament compiler infrastructure. I don't think it is specifically architected to contained truly agnostic utilities.

[UnsignedByte]

Macro-ized a bunch of stuff but I'm holding off on adding port attributes for another pass.

[rachitnigam]

Any reason to expose the comp_attrs submodule?

[UnsignedByte]

Technically each set of Attrs comes with three types: Bool, Num, and Attrs for boolean and numeric flags as well as a type alias for the struct with these two types of flags.

We could publicize the underlying types instead, but that would then look something like CompAttrs, CompBool, CompNum and PortAttrs, PortBool, PortNum (for port attributes). Matter of preference honestly, I don't really mind either way.