Introduce __smo intrinsic

[mohammadfawaz]

Closes #3457

Following how __log works for the most part.

Tasks:

• Parser
• Frontend
• JSON ABI (following Add messagesTypes to the JSON ABI fuel-specs#444)
• IR
• IR Gen
  • Unit tests
• Codegen
  • Unit tests
• Standard library change
  • SDK test
• E2E tests
• Documentation for __smo

Nothing particularly special about this change except for IR gen. The new intrinsic has the following signature:

__smo<T>(recipient: b256, data: T, output_index: u64, coins: u64)

while the IR instruction has the following signature:

smo recipient_and_message, message_size, output_index, coins

which matches the smo opcode. IR gen builds recipient_and_message by creating a new struct where the first field is the recipient (a b256), the second field is the smo ID (just like log IDs, but we don't have a special register to use here so we need to use the message data), and the third field is the actual user data passed in via the intrinsic:

struct recipient_and_message {
    recipient: b256,
    smo_id: u64,
    user_data: T
}

The smo_id above is the same as the one used in the JSON ABI for each smo.

Assembly builder simply convert the IR instruction into an smo opcode directly.

Finally, add a send_typed_message to the standard library that allows sending arbitrary types and added an SDK test for it. The SDK can't decode the types yet so I had to check for raw bytes.

[mohammadfawaz]

Sorry for the trivial JSON ABI file diffs, again

[otrho]

Do we need to add an SDK test to make sure the message actually gets through correctly? Is that possible?

[vaivaswatha]

Looks fine to me (after Toby's comments). So approving in advance.

[mohammadfawaz]

Do we need to add an SDK test to make sure the message actually gets through correctly? Is that possible?

Likely better to do it here so I did. I originally thought it can be done in #3577 but might as well get it done.

[nfurfaro]

I just left 1 question but looks great overall!