Improve cost estimation/calculation if batch job was cancelled/failed

[zansinergise]

We should investigate how much costs were actually incurred if a batch job was cancelled or it failed during processing, since the costs/usage would probably be then lower than what we estimated before starting the job.

A bit more context: https://git.sinergise.com/team-6/openeo-sentinelhub-python-driver/-/merge_requests/318#note_583133

[zcernigoj]

Cost of a failed tile inside a SH batch job is 0, for Sentinel Hub (platform itself).

Consequently

• the cost of SH batch job where all tiles failed, is 0
• the cost of SH batch job where some tiles failed, is lower

Can't find it in https://docs.sentinel-hub.com/api/latest/api/batch/
But can be checked for failed and partial SH batch jobs in SH dashboard for app_admin_openeo account.

---

On a separate note: we should determine, what the fixed (same for all jobs - failing or not) and variable (different for successful and failing jobs) costs are on AWS (lambda+dynamodb+s3) for the driver