Improve Dispatch Errors

[echevrier]

For now we only decode the module variant and create a module error. All other errors are not decoded.
The Balances' calls throw now TokenError , which are not module errors and so not decoded.

subxt did it in a nice way, using scale-decode(https://github.com/paritytech/scale-decode) and scale-encode : paritytech/subxt#878
If we want to go this way we have to make these libs no-std compatible.

[haerdib]

It seems scale value has been split into 3 crates: scale-value, scale-encode and scale-decode.

e.g. scale-value decode now uses the fairly similar function of scale-value:
https://github.com/paritytech/scale-decode/blob/main/scale-decode/src/visitor/decode.rs#L32-L61

new scale value:
https://github.com/paritytech/scale-value/blob/main/src/scale_impls/decode.rs#L26-L32

old scale value:

pub fn decode_value_as_type<Id: Into<TypeId>>(
	data: &mut &[u8],
	ty_id: Id,
	types: &PortableRegistry,
) -> Result<Value<TypeId>, DecodeError> {
	let ty_id = ty_id.into();
	let ty = types
		.resolve(ty_id.id())
		.ok_or_else(|| DecodeError::TypeIdNotFound(ty_id.id()))?;

	let value = match ty.type_def() {
		TypeDef::Composite(inner) =>
			decode_composite_value(data, inner, types).map(ValueDef::Composite),
		TypeDef::Sequence(inner) =>
			decode_sequence_value(data, inner, types).map(ValueDef::Composite),
		TypeDef::Array(inner) => decode_array_value(data, inner, types).map(ValueDef::Composite),
		TypeDef::Tuple(inner) => decode_tuple_value(data, inner, types).map(ValueDef::Composite),
		TypeDef::Variant(inner) => decode_variant_value(data, inner, types).map(ValueDef::Variant),
		TypeDef::Primitive(inner) => decode_primitive_value(data, inner).map(ValueDef::Primitive),
		TypeDef::Compact(inner) => decode_compact_value(data, inner, types),
		TypeDef::BitSequence(inner) =>
			decode_bit_sequence_value(data, inner, types).map(ValueDef::BitSequence),
	}?;

	Ok(Value { value, context: ty_id })
}

But using the visitor seems to allow zero-copy decoding, which is a huge benefit. But I'm not sure how easily that's convertable into no-std. We'll have to look into that. If that's not possible, I think we can continue using the old version (with maybe slight updates).

[echevrier]

Convert it into no-std requires to convert:

• https://github.com/paritytech/scale-decode
• https://github.com/paritytech/scale-bits
• https://github.com/paritytech/scale-encode

They use thiserror, which is std-only.

[echevrier]

For now we add a quick fix without reworking all the encoding/decoding of substarte-api-client. This should be done in another step. See #564