Introduce precise fallback decompiler configuration

[sifislag]

The main item of this PR is the introduction of the precise fallback decompilation configuration option. When this option is enabled, if the default decompilation configuration (transactional) completes with imprecision (non empty Analytics_JumpToMany) it attempts to re-decompile it with a more precise configuration (currently employs the same kind of block cloning as the --early_cloning flag).

Precise fallback is on by default and can be disabled using the --disable_precise_fallback flag. In addition the --single_decomp flag has been removed and --disable_scalable_fallback has been introduced.

Other items:

• refactor timeouts logic:
  • distinguish decompilation and client timeouts in logs and json results.
  • handle the case where the process memory limit is exceeded (only for decompilation)
• change default minimum client time to 10s.
• cleanup of dl code
• added test cases for scalable_fallback and precise_fallback
• fixed regression on some unoptimized memory patterns produced by abiencodewithselector

[sifislag]

For 1000 contracts for the elipmoc dataset (default is with the scalable fallback):

Timeouts are unchanged
Total decompilation time:
default: 6031 s
default+precise_fallback: 9826 s

JumpToMany:
default: 591
default+precise_fallback: 364

NonModeledMLOAD:
default 16718
default+precise_fallback: 14189

[sifislag]

For 1000 contracts for the elipmoc dataset without the scalable fallback config:

Timeouts are unchanged
Total decompilation time:
transactional: 3118 s
transactional+precise_fallback: 6938 s

JumpToMany:
transactional: 403
transactional+precise_fallback: 176

NonModeledMLOAD:
transactional: 15340
transactional+precise_fallback: 12831