intrinsic-test: Implemented DerefMut for ArmIntrinsicTest

[madhav-madhusoodanan]

Summary

1. Moved the return type of IntrinsicType::from_c to Rust<Self, String> from Result<Box, String>
2. Implemented DerefMut for ArmIntrinsicTest struct

Context

This PR is part 2 of the changes that were originally made in the x86 extension PR #1814 for intrinsic-test, and is intended to unblock the speedup efforts that are being made in PRs #1856 , #1862 and #1863.

[folkertdev]

Is this really worth it?

In general it is my impression that this code base is much too generic already (dyn, bunch of traits), but then only had one concrete implementation.

Deref is generally discouraged, why does it pull its weight here?

[madhav-madhusoodanan]

There are 3 levels of functionality for structs that implement IntrinsicTypeDefinition:

1. The basics (name, is_simd, is_ptr, etc) where the logic is strightforward and doesn't need to be implemented per architecture
2. The mid level (get_load_function, c_type, etc) where the logic may vary across architectures, and would need architecture-specific definition
3. The code generation part (generate_c_test_loop, format_c_main_template, etc) where the logic is again common across architectures:
   a. Define headers
   b. Define any typecasting functionality as per necessity
   c. Define the variable
   d. Define the list of values that would be passed as an argument to the intrinsic
   e. Define test loops
   f. Load the arguments, call the intrinsics and print the value

We figured that:

1. Any functionality that would be needed at Level 2 could use Deref and cleanly use the functions defined on the inner type.
2. Any functionality that would be needed at Level 3 could use Generics and associated traits to cleanly use the architecture-specific functionality

[madhav-madhusoodanan]

I look forward to learning how we can better design intrinsic-test so that it supports easy addition of other architectures too.