Support MDF writing in HLT2

We should support persisting all information from HLT2 in a single LHCb::RawEvent object. The production use-case requires writing a single raw event per stream, as does writing out an MDF.

Today Moore only supports writing a DST when running HLT2. This is because the Turbo-like persistence saves packed containers rather than raw banks. Once the latter is available, we can concatenate the other raw banks already created (luminosity and various reports) into /Event/DAQ/RawEvent or wherever it should go.

---

Somewhat of an aside: we could consider supporting both packed container and raw bank outputs all them DST-like and MDF-like). The benefit of DST-like is that one does not necessarily need TCK-like information to read the file, whereas you do for MDF-like (to decode the raw banks). Even for DST-like you do need a TCK for decode the reports banks, but not every check requires this (e.g. Moore has several test that just unpack the Turbo information).