Add internal hash map of literal indexes:
In this hash map key is pair of block number and literal's unique identifier in the block, and the value is a literal index that unique in the whole program.
Block is a continues array of opcodes. So, bytecode is splitted into blocks.
Each block has its own uid counter. To get literal index the interpreter looks up it in the hash map.
Thus, now JS program is able to have more than 255 identifiers/string literals.
The first 128 (0-127) uids are reserved for block's uid counter, the other 128 (128-255) are reserved for tmp variables.
Fix varg generation: generate *_n opcodes with parameters in following meta opcodes
Add stack internal structure: dimanically allocated stack.
Use dynamically allocated memory in parser: every local and global variables are stored in dinamically allocated stacks.
Use dynamically allocated memory in serializer: opcodes are also stored in stack.
Change is_true_jmp and is_false_jmp opcodes to relative.
Change *jmp* opcodes to be able to store opcode_counter_t instead of idx_t.
