Make the search start from the best reward position

[nicolaefilat]

Changes I made:

• Add select that selects top 5 best programs in terms of reward
• Add update_grammar for traces that is not used yet. It updates based on reward instead of examples.
• Refactor x_player_start variable (and the other y,z) to be a float and not a float array.
• Add new_action["chat"] = "/effect @a minecraft:instant_health 1 10000 true" for each action such that the player can't die :)
• Made negative rewards be "equivalent" to each other such that they will be not added to the bank.
• When the synthesis restarts, the player will start from the best position found so far. This is a heuristic that seems to work

[codecov]

Codecov Report

Attention: Patch coverage is 5.96026% with 142 lines in your changes are missing coverage. Please review.

Please upload report for BASE (probe-with-minerl@bc0079a). Learn more about missing BASE report.

Files	Patch %	Lines
src/probe/probe_iterator.jl	12.24%	43 Missing ⚠️
src/minecraft/getting_started_minerl.jl	0.00%	34 Missing ⚠️
src/minecraft/create_minerl_env.jl	0.00%	33 Missing ⚠️
src/probe/update_grammar.jl	0.00%	24 Missing ⚠️
src/probe/guided_trace_search_iterator.jl	0.00%	8 Missing ⚠️

Additional details and impacted files

@@                 Coverage Diff                  @@
##             probe-with-minerl     #106   +/-   ##
====================================================
  Coverage                     ?   60.56%           
====================================================
  Files                        ?       32           
  Lines                        ?     1202           
  Branches                     ?        0           
====================================================
  Hits                         ?      728           
  Misses                       ?      474           
  Partials                     ?        0           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[nicolaefilat]

Ok is good