Refactor XCM code

[xlc]

Currently XCM code are not implement in an extendable way. To make a new version, we have to duplicate many files and then modify on the duplicated files. This comes with many issues:

• A lot of work to add a new version
• Impossible to tell difference between two versions by only looking at the code
• Duplicated code is bad in general

We should refactor it in such way to avoid duplications.

For example, instead of a bigenum Instruction, we have a struct for each instruction and use some macro to generation the big enum. In that way, to add new instruction, we only need to add new instruction and update the instruction list.

[acatangiu]

Added "good-first-issue" label for the remaining work in #7123

Most of the heavy-lifting there is done, but we need someone to drive it over the finish line. Contributions are welcome, we can help with guidance and reviews.