feat!: split candid crate

[lwshang]

Currently, the candid crate mainly consists of two parts:

• Candid data encoding/decoding
• Parser and derived functionality (bindgen)

The second part evolves fast which forces us to bump major version frequently.
Such version bumps propagated to the downstream projects (agent-rs, cdk-rs, sdk, ic monorepo, etc).

By splitting out candid_parser, we expect to have a stable candid crate.

Also, I move Principal into a separate ic_principal crate. Therefore, we can provide a lightweight dependency for those who only need Principal.

Notable changes

• Some private or crate public methods are now public since they are called cross-crate
• Add parse_idl_args() and parse_idl_value() because in candid_parser we cannot implement FromStr for IDLArgs and IDLValue which are defined in candid.
• TypeEnv is defined in candid, so we can't implement type method ast_to_type in candid_parser. Instead I turned it into a function which takes TypeEnv as the first argument. Similarly, size_helper() and size() are changed.

Checklist

• Overhaul the Cargo.toml files
• Changelog

[chenyan-dfinity]

Mostly LG. Can you also update the changelog?

[github-actions]

Benchmark for c99ab8d

Click to view benchmark

Test	Base	PR	%
Blob/&str	245.4±70.88µs	247.0±70.86µs	+0.65%
Blob/ByteBuf	158.0±46.41µs	155.2±47.27µs	-1.77%
Blob/Bytes	112.7±43.36µs	113.8±44.02µs	+0.98%
Blob/String	319.3±150.88µs	322.0±152.73µs	+0.85%
Collections/vec (text, nat)	70.4±2.07ms	70.2±1.44ms	-0.28%
Collections/vec int	32.2±0.09ms	32.2±0.09ms	0.00%
Collections/vec int64	19.2±0.31ms	18.7±0.30ms	-2.60%
Collections/vec nat8	14.3±0.03ms	14.7±0.14ms	+2.80%
option list/1024	1378.9±1.80µs	1381.9±2.69µs	+0.22%
profiles/1024	2.7±0.00ms	2.7±0.02ms	0.00%
variant list/1024	1121.5±2.52µs	1119.2±4.94µs	-0.21%

[chenyan-dfinity]

Let's merge this to a beta branch. I have a few other breaking changes to make after this. Let's aim for an official release early November.

[chenyan-dfinity]

Do you recall why we put Principal back to candid a while ago? Also do we want to give it a more general name, e.g. ic_types, to accommodate more types in the future?

[lwshang]

We used to have the ic-types crate which contained principal and hash_tree modules.
The modules were moved to candid and agent-rs repo respectively. And ic-types is deprecated.

IMO, it's better to have ic_principal crate which contains only Principal. It will be a stable crate on the top of the dependency tree. We won't have to bump its version because of changes in other parts.

[lwshang]

I just pushed the beta branch which is identical to this lwshang/split branch.

[github-actions]

Benchmark for 94c03cc

Click to view benchmark

Test	Base	PR	%
Blob/&str	262.0±84.31µs	263.1±88.02µs	+0.42%
Blob/ByteBuf	232.5±76.58µs	223.1±55.06µs	-4.04%
Blob/Bytes	151.0±58.76µs	160.8±57.55µs	+6.49%
Blob/String	362.3±158.23µs	367.2±167.97µs	+1.35%
Collections/vec (text, nat)	82.3±3.56ms	82.4±4.10ms	+0.12%
Collections/vec int	36.2±1.65ms	37.3±2.24ms	+3.04%
Collections/vec int64	20.8±0.98ms	21.4±1.30ms	+2.88%
Collections/vec nat8	16.2±0.61ms	15.7±0.51ms	-3.09%
option list/1024	1624.1±86.29µs	1639.4±76.59µs	+0.94%
profiles/1024	3.1±0.15ms	3.1±0.13ms	0.00%
variant list/1024	1280.4±61.30µs	1324.7±110.38µs	+3.46%