Apromore no longer requires the presence of "event_nr" in logs.

[raboczi]

Change to collate-events.py reduces the requirements on input logs to:

• traces have concept:name (becomes case_id)
• events have time:timestamp (used within the ML scripts)

I believe both of these are guaranteed simply by being expressed in XES.

Change to predict.py downloads pickles from Apromore directly into memory, rather than requiring a temp file.

Change to README.md should now format Windows batch file code snippets correctly.