Spike dmd codelist builder

[Jongmassey]

Fixes #176

In-person hack with @evansd and @inglesp to provide ingredient, VMP, and AMP name based building of dm+d codelists

[Jongmassey]

Having done a bit more digging, the NHSBSA dm+d supplementary file (aka dmdbonus) has VTM to ingredient mapping. This presents us with an opportunity to present the search results by VTM hierarchy more clearly.

This would require the dmd supplementary data to be added to the dm+d coding system.

For example, the descendent relationships defined here, go from ing-to-vmp, ing-to-amp, vmp-to-amp. This could be simplified to (assuming vtm ingredient as VTI)

vtms = VTM.objects.using(self.database_alias).filter(
   Q(vti__ing_id__in=codes) | Q(id__in=codes)
)
vmps = VMP.objects.using(self.database_alias).filter(
   Q(vtm__in=[vtm.id from vtm in vtms_from_ings]) | Q(id__in=codes)
)
amps = AMP.objects.using(self.database_alias).filter(vmp_id__in=[vmp.id for vmp in vmps_from_ings])

rv = (
   {
      (vti.ing_id, vtm.id)
      for vtm in vtms 
      for vti in vtm.vti_set.all()
    }
    | {(vmp.vtm_id, vmp.id) for vmp in vmps}
    | {(amp.vmp_id, amp.id) for amp in amps}
  )
return rv