fix(EOF): panic on empty input range, and continue exec after eofcreate

[rakita]

InstructionResult needs to be reset when eofcreate continues execution.

And Handle panic on empty range

[github-actions]

Valgrind Results:

==3776== Cachegrind, a cache and branch-prediction profiler
==3776== Copyright (C) 2002-2017, and GNU GPL'd, by Nicholas Nethercote et al.
==3776== Using Valgrind-3.18.1 and LibVEX; rerun with -h for copyright info
==3776== Command: target/release/snailtracer
==3776== 
--3776-- warning: L3 cache found, using its data for the LL simulation.
Running snailtracer example!
elapsed: 1.810893088s
==3776== 
==3776== I   refs:      429,375,370
==3776== I1  misses:          3,142
==3776== LLi misses:          2,802
==3776== I1  miss rate:        0.00%
==3776== LLi miss rate:        0.00%
==3776== 
==3776== D   refs:      206,206,144  (133,651,464 rd   + 72,554,680 wr)
==3776== D1  misses:        345,426  (    205,298 rd   +    140,128 wr)
==3776== LLd misses:        137,725  (      4,326 rd   +    133,399 wr)
==3776== D1  miss rate:         0.2% (        0.2%     +        0.2%  )
==3776== LLd miss rate:         0.1% (        0.0%     +        0.2%  )
==3776== 
==3776== LL refs:           348,568  (    208,440 rd   +    140,128 wr)
==3776== LL misses:         140,527  (      7,128 rd   +    133,399 wr)
==3776== LL miss rate:          0.0% (        0.0%     +        0.2%  )

[rakita]

Note I assumed that there is wrong order of stack pop's for EXTCALL, but spec says that original impl was correct, and order is not same as CALL opcode. Order is: EXTCALL (0xf8) with arguments (target_address, input_offset, input_size, value)