I/O for decayed events

[kilian]

Plain rereading of decayed events (raw) doesn't work as one would expect - event transforms are neglected, and analysis of decay products therefore can't be reproduced. The decay chain could be recreated from random numbers, but this is not exactly the caching mechanism that the raw format should provide.

The raw I/O format is insufficient, see also #627.

I don't think that this worked before, so it is a actually new feature. Obviously related to #663 and others.

[kilian]

See also #608.

[kilian]

Included a test case rebuild_5 for this issue (which fails, as expected). r6510

[kilian]

r6512: test case rebuild_5 now works, i.e., dressed events are cached in the .evx file and can be analyzed in a subsequent run.

Next, check other situations and look at lhef_7 and 8.

[kilian]

lhef_7 and lhef_8 are fixed in r6513. This was more related to #663, i.e., the meaning of update_event.