Reduce evt_max for allen_mdf_write test

I've observed quite a few times that the allen_mdf_write test runs into timeouts even on the opt platforms.
e.g.: v2_opt, v3_opt, v2_clang
I assume it's because I/O is quite a fought over resource in the nightlies.

This reduces the amount of events from 1k -> 333.

@dovombru @dcampora I'd be interested to hear your thoughts if this is ok or if there are any objections from your side to this reduction?