EVM runtime: Map FVM context info to EVM opcodes

[karim-agha]

See #689 (comment) for the updated view and state

---

Implement opcodes that query the transaction context:

• ORIGIN
• CALLVALUE
• ADDRESS
• BLOCK_NUMBER
• TIMESTAMP
• CALLER
• BLOCKHASH
• CHAINID
• BASEFEE
• DIFFICULTY
• GASLIMIT

[raulk]

These should be simple, except for the address ones, where we'll eventually need to return the f4 address.

[raulk]

Notes / checklist.

Done

• ORIGIN => requires a new syscall: Add a syscall to obtain the originator of a call #300.
• CALLVALUE => maps to the value received in the transaction context, in native FIL.
• ADDRESS => returns the ID address in EVM form of the receiver.
• BLOCK_NUMBER => epoch.
• CALLER => returns the ID address in EVM form of the caller.
• BASEFEE => return the basefee
• DIFFICULTY => return 0
• SELFBALANCE => return balance of the current contract

Requires FVM work:

• GASPRICE => sum of base_fee and gas_premium. Requires the gas_premium in the InvocationContext, so that it's returned in vm::context().
• BALANCE => requires a syscall to obtain the balance of an arbitrary address.
• GAS => return available gas. Requires a gas::available() syscall.

** Requires FVM, ffi, and client work: **

• TIMESTAMP => the timestamp is an attribute in the block header; needs a syscall and snake through the ffi from the client.
• BLOCKHASH => this is tricky. I don't think we can return just the hash of the first block in the parent tipset, as we probably lose some security in the face of a reorg that preserves the first block.
• CHAINID => return value zero until Chainlist submission to reserve a chain ID #681?
• GASLIMIT => returns the block's gas limit. We may need a syscall or a context parameter to return this value to avoid hardcoding the 15B gas assumption.

For the latter group consider:

1. Putting these attributes behind a chain_data() syscall. These are really "environmental attributes" not scoped to the message (unlike InvocationContext).
2. Adding them to the InvocationContext returned by vm::context(), but that would bloat the struct for cases that do not need to access all attributes (likely not a big deal, it'll be some ~64 bytes extra)

[raulk]

Now tracking more granularly in #864 and #865.