Write tests for bugs caught while syncing the chain

[SamWilsn]

Originally posted by @lightclient in #373 (review)

Should re-check to see if these have been added to EEST since this issue was written

This is a bit of a meta-issue for tracking bugs that weren't caught by the tests that were triggered while syncing the chain.

These tests should theoretically get upstreamed to https://github.com/ethereum/tests and checked off here once they're merged.

• sstore persists between transactions (Fix optimized state bugs #368)
• charge gas for ECRECOVER
• blockhash should hash only the header, not the block (Fix calculation of previous block hash #373)
• signextend (Sync fixes #375)
• Push-data extends past end of contract code (Sync fixes #375)
• Deploying an empty contract (Sync fixes #375)
• ECRECOVER with truncated call data (ecrecover doesn't handle truncated calldata correctly #445)
• Logs emitted during CREATE2 #466
• Blockhash at 255/256/257

[petertdavies]

Here are some more edgecases that are untested in legacytests. We shouldn't implement tests until the specs support the non-legacy tests (where stuff is likely to be tested).

• Check compliance with condition d) in EIP 161
• If the modulus sent to the modexp precompile is zero, the return value is zero of the same length as the modulus. The "same length" condition is untested.
• A block is invalid if it contains a transaction which the sender cannot afford.
• The difficulty adjustment is not integration tested as part of a larger block. This caused a bug in a version of [WIP] Verkle changes to test t8n execution-spec-tests#465 to not get caught by the test suite.
• Log ordering is not tested.
• MSTORE8 with value 0x0 is not tested.
• Returndata clearing is not tested in failure cases of CREATE/CREATE2

[gurukamath]

• Check compliance with EIP-3607 (EIP-3607: Reject transactions from senders with deployed code #583)

[SamWilsn]

#1497 would probably be useful for this

[gurukamath]

• Also needd to upstream the insufficient_balance_for_value test case that we wrote some time ago, if not done already. See this