pruntime: Detect checkpoint breaking changes in unittest

[kvinwang]

This PR adds #[derive(TypeInfo)] onto the pruntime checkpoint data types and dump the type tree in unittest.
Ideally, it would trigger a test failure once there are some changes in the type tree.
However, it is not 100% accurate, becuase the different behaivor between serde and TypeInfo. For example, TypeInfo always skips zero sized types but serde doesn't.

[h4x3rotab]

Brilliant idea. How do you come up with the TypeInfo idea? Also, codec(skip) is applied for most of the serde(skip) fields, but not all. How do you choose them?

[kvinwang]

Brilliant idea. How do you come up with the TypeInfo idea?

In PR #1363, a breaking change was introduced. I came up with this solution while troubleshooting the issue.

Also, codec(skip) is applied for most of the serde(skip) fields, but not all. How do you choose them?

The purpose of #[codec(skip)] is to skip fields that are hard to implement TypeInfo for; And our mind are aware of these fields would unlikely change.

[h4x3rotab]

lgtm