Avoid PUSH data when computing jumpdests

[anvacaru]

Fixing a mistake I made in my previous PR #2441. When computing the valid JUMPTESTS, I omitted the case in which the 0x5B byte could have been part of a PUSH(N) opcode.

I've modified the _compute_jumpdests function to first check if the current byte is a PUSH(N) and skip byte - PUSH1 +2 bytes.

i.e. if byte is PUSH1, it skips 0x60 - 0x60 + 2 bytes where one byte is the data to be pushed and one byte is the opcode itself.