uni code bloat

[matklad]

We have this snippet of code in the sdk right now:

near-sdk-rs/near-sdk/src/json_types/public_key.rs

Lines 27 to 31 in 763e23c

	match value.to_lowercase().as_str() {
	"ed25519" => Ok(CurveType::ED25519),
	"secp256k1" => Ok(CurveType::SECP256K1),
	_ => Err(ParsePublicKeyError { kind: ParsePublicKeyErrorKind::UnknownCurve }),
	}

It has two problems. The minor one is that to_lowercase allocates a string, while we can do this comparison without allocations.

The bigger problem is that .to_lowercase() does unicode-aware case conversion. Such conversion requires a sizable unicode table:

https://github.com/rust-lang/rust/blob/5249414809d40fe22eca0c36105a2f71b9006e04/library/core/src/unicode/unicode_data.rs#L576-L1386

This table is a part of libcore, and gets linked with every contract. It's not possible to opt-out of unicode support (which is sort-of a design bug for core), but it can be eliminated by the linker if it isn't actually used. Naturally, .to_lowercase() uses the table and prevents the linker from eliminating it.

We should do the following:

• write some kind of a test that demonstrates that today contracts include this table (probably need to actually call this function)
• replace to_lowercase with [eq_ignore_ascii_case](https://doc.rust-lang.org/stable/std/primitive.str.html#method.eq_ignore_ascii_case)
• verify with the test that the table goes away
• if it doesn't go away, scrutinize sdk for usages of other unicode-aware methods.

[austinabell]

This should be removed, but not going to close the issue yet because all code paths of the SDK haven't been tested that this isn't included