fix(db): Set instruction result at outcome insert

[rakita]

self.instruction_result = InstructionResult::Continue; was set always before running the interpreter loop to override the InstructionResult::CallOrCreate but it would void the Fatal result that was set in outcome insert.

So moving override to outcome insert fixes it.

[github-actions]

Valgrind Results:

==3988== Cachegrind, a cache and branch-prediction profiler
==3988== Copyright (C) 2002-2017, and GNU GPL'd, by Nicholas Nethercote et al.
==3988== Using Valgrind-3.18.1 and LibVEX; rerun with -h for copyright info
==3988== Command: target/release/snailtracer
==3988== 
--3988-- warning: L3 cache found, using its data for the LL simulation.
Snailtracer Host+Interpreter benchmark (2.1s) ...            not enough samples
==3988== 
==3988== I   refs:      471,785,698
==3988== I1  misses:          7,163
==3988== LLi misses:          2,917
==3988== I1  miss rate:        0.00%
==3988== LLi miss rate:        0.00%
==3988== 
==3988== D   refs:      229,164,273  (152,345,985 rd   + 76,818,288 wr)
==3988== D1  misses:        350,871  (    210,583 rd   +    140,288 wr)
==3988== LLd misses:        137,578  (      4,054 rd   +    133,524 wr)
==3988== D1  miss rate:         0.2% (        0.1%     +        0.2%  )
==3988== LLd miss rate:         0.1% (        0.0%     +        0.2%  )
==3988== 
==3988== LL refs:           358,034  (    217,746 rd   +    140,288 wr)
==3988== LL misses:         140,495  (      6,971 rd   +    133,524 wr)
==3988== LL miss rate:          0.0% (        0.0%     +        0.2%  )