fect(revme): EmptyDb Blockhash string, json-outcome flag, set prevrandao in statetest

[rakita]

Noted here: holiman/goevmlab#113 (comment)

• EmptyDb blockhash was returning kecak of a number while other clients are returning kecak of the string of that number
• Added --json-outcome flag that would printout the stateRoot after execution, it is implied if --json is enabled, and it will stop second execution of test.
• Prevrandao field from json test was not used, we used difficulty it was probably same value inside eth/tests. So corrected this.

[github-actions]

Valgrind Results:

==3793== Cachegrind, a cache and branch-prediction profiler
==3793== Copyright (C) 2002-2017, and GNU GPL'd, by Nicholas Nethercote et al.
==3793== Using Valgrind-3.18.1 and LibVEX; rerun with -h for copyright info
==3793== Command: target/release/snailtracer
==3793== 
--3793-- warning: L3 cache found, using its data for the LL simulation.
Snailtracer Host+Interpreter benchmark (2.1s) ...            not enough samples
==3793== 
==3793== I   refs:      455,350,780
==3793== I1  misses:         16,085
==3793== LLi misses:          2,878
==3793== I1  miss rate:        0.00%
==3793== LLi miss rate:        0.00%
==3793== 
==3793== D   refs:      237,389,298  (160,565,586 rd   + 76,823,712 wr)
==3793== D1  misses:        349,540  (    209,231 rd   +    140,309 wr)
==3793== LLd misses:        137,566  (      4,052 rd   +    133,514 wr)
==3793== D1  miss rate:         0.1% (        0.1%     +        0.2%  )
==3793== LLd miss rate:         0.1% (        0.0%     +        0.2%  )
==3793== 
==3793== LL refs:           365,625  (    225,316 rd   +    140,309 wr)
==3793== LL misses:         140,444  (      6,930 rd   +    133,514 wr)
==3793== LL miss rate:          0.0% (        0.0%     +        0.2%  )