Non-conforming command sets raise an exception

[or150]

Describe the bug
Hi,

I am working with a PACS that has a weird implementation bug. It returns two NumberOf*Suboperations fields in C-Move responses.
When parsing the command set, pydicom parses it as MultiValue, and this in turn causes pydicom to raise:
TypeError: Number of Remaining Suboperations must be an int

Expected behavior
We are trying to migrate from dcmtk/dcm4che3 cli utilities, and encountered this bug only here.

From looking at both implementations, it seems that dcmtk takes the first value:
https://github.com/DCMTK/dcmtk/blob/a9561f364d50adf5344d45e01fda0b2ae784877f/dcmnet/libsrc/dimcmd.cc#L1044
https://github.com/DCMTK/dcmtk/blob/a9561f364d50adf5344d45e01fda0b2ae784877f/dcmnet/libsrc/dimcmd.cc#L280

And if I didn't miss anything, dcm4che3 seems to completely ignore it (checks only the Status field, and takes the first value there, too):
https://github.com/dcm4che/dcm4che/blob/5b3f94cc8816490e6e0cf5dc82516901b828fecf/dcm4che-tool/dcm4che-tool-movescu/src/main/java/org/dcm4che3/tool/movescu/MoveSCU.java#L357

Since both are popular implementations, I assume that taking the first value is ok for these cases.

I have a fork that before setting any field of a DIMSEPrimitive from the command dataset, verifies the VM of the tag against dictionary_VM and takes the first value if needed. If this solution is acceptable, I can open a PR.

Steps To Reproduce

from pynetdicom.pdu_primitives import P_DATA
from pynetdicom.dimse_messages import C_MOVE_RSP

def reproduce():
    cmd_set = b'\x03\x00\x00\x00\x00\x04\x00\x00\x00\x72' \
              b'\x00\x00\x00\x00\x00\x02\x00\x1a\x00\x00' \
              b'\x00\x31\x2e\x32\x2e\x38\x34\x30\x2e\x31' \
              b'\x30\x30\x30\x38\x2e\x35\x2e\x31\x2e\x34' \
              b'\x2e\x31\x2e\x31\x2e\x32\x00\x00\x00\x00' \
              b'\x01\x02\x00\x00\x00\x21\x80\x00\x00\x20' \
              b'\x01\x02\x00\x00\x00\x05\x00\x00\x00\x00' \
              b'\x08\x02\x00\x00\x00\x01\x00\x00\x00\x00' \
              b'\x09\x02\x00\x00\x00\x00\xff\x00\x00\x20' \
              b'\x10\x04\x00\x00\x00\x03\x00\x00\x00\x00' \
              b'\x00\x21\x10\x04\x00\x00\x00\x01\x00\x00' \
              b'\x00\x00\x00\x22\x10\x04\x00\x00\x00\x02' \
              b'\x00\x00\x00\x00\x00\x23\x10\x04\x00\x00' \
              b'\x00\x04\x00\x00\x00'

    data_set = b'\x02\x08\x00\x52\x00\x08\x00\x00\x00\x50' \
               b'\x41\x54\x49\x45\x4e\x54\x20\x10\x00\x20' \
               b'\x00\x02\x00\x00\x00\x2a\x20'

    msg = C_MOVE_RSP()
    for data in [cmd_set, data_set]:
        p_data = P_DATA()
        p_data.presentation_data_value_list.append([0, data])
        msg.decode_msg(p_data)
    primitive = msg.message_to_primitive()


reproduce()

Your environment
Almost Latest pynetdicom development code (93ef29c)
pydicom 2.0.0
python 3.7

[scaramallion]

It's probably easier to assume the VM should be 1 as only Offending Element and Attribute Identifier List are 1-n

This should be faster than the dictionary_VM lookup

for elem in self.command_set:
    if hasattr(primitive, elem.keyword):
        if elem.VM > 1 and elem.tag not in [0x00000901, 0x00001005]:
            LOGGER.warning(f"Non-conformant VM {elem.VM} for '{elem.keyword}', taking the first value")
            elem.value = elem.value[0]
        setattr(primitive, elem.keyword, elem.value)

If you want to submit a PR that'd be great, otherwise I'm happy to do it.

[or150]

Opened a PR.
One of the tests broke, and I'm not sure that the that case is valid anymore.
I'd be glad to get feedback about it