`MaybeResize` for `Uint`

I open this as an issue as I'm sure it'll need discussion.

Problem:

I need 1 << x where x is ~256-2000, but variable and specified by the caller. I want either:

• The caller to pass me T = U2048
• The caller to pass me T = BoxedUint

My issue is if I get T = BoxedUint, I cannot get a 2048-bit BoxedUint without using BoxedUint::one_with_precision, Resize, or BoxedUint::one_like. The first two are specific to BoxedUint. The third is portable with Uint, but would require the caller pass in one_shl_x so that I had the value to 'be like'.

Ideally, I'd have a MaybeResize::resize(One::one(), x) which for fixed types, does nothing, but for BoxedUint, does call BoxedUint::resize.

Are we fine with MaybeResize as a complement to Resize for this purpose? Is there some function I'm missing? Should we instead have One::one_with_precision?

As an absolute hack, I can simply have an RNG which always outputs 0 and abuse RandomBits for this today. As an oddity, RandomBits::try_random_bits_with_precision appears to be the only trait method today which is implemented for both Uint and BoxedUint, while supporting a precision argument (which is discarded if inapplicable). That at least suggests precedent for this...

Read the existing One, Resize, and RandomBits trait APIs, especially BoxedUint::resize and RandomBits::try_random_bits_with_precision, to understand current fixed- versus variable-precision behavior. Done means the project has agreed on a portable way to construct a one value at caller-selected precision, with its scope and trait API settled.