Metrics for benchmarks

[jmatejcz]

Come up with a metrics that will calculate how accurate the task was done.
After couple specific implementations think of common instructions/tools that might be useful when creating a metric

[jmatejcz]

• add verification to tasks, the verification will ensure that the scene is suitable for the task, in case when it is not suitable(for ex. move carrot, but there is no carrot in the scene) it should not be possible to create such scenario
• add a way for creating scenarios automatically from given list of tasks and scenes

[jmatejcz]

• define tasks more clearly
• ensure proper naming and calculations

[jmatejcz]

• number of tool calls?
• calculating score also at the begining
• storing results

[jmatejcz]

• ensure all required service are running before starting scenario

[jmatejcz]

• proper log after finishing all scenarios
• logging how many scenarios left

[jmatejcz]

• dump result into file

[jmatejcz]

• starting every scenario with score 0.0
• mark every task by the difficulty (subjective? based on number of moves required?) - instead grouped scenarios into packets of difficulty - feat: add more scenarios to o3de manipulation benchmark #452

[jmatejcz]

introduced in #436
then extended in #444 and #452