291 replace dicom anonymization

[peshence]

291 Switch from DIY DICOM anonymiasation to Kitware DICOM Anonymizer. In the process:

• Public DICOM tags now have default actions as per the library.
• old "hash" function applied to UIDs while keeping their prefix has been replaced with a random UID replacement (no prefix continuity) as we don't currently believe loading images on original scanner is likely.
• only patientID is now hashed via hasher api (as patientID+Acc_num). Previously Patient name and acc_num were also hased like this, now they are replaced as empty strings in existing projects since they shouldn't be used for linking.

Possible improvement: Convert current YAMl tag configs to JSON to make use the the existing underlying action loading functionality. We can then do regex replacements and a few other slightly more complex actions. Not sure if it's necessary at this stage.

Possible further actions, which I think shouldn't be in this PR:

• fix imaging tests path mounting to work regardless of PWD
• make EHR API use only single hasher endpoint and remove extra endpoints from hasher api

[milanmlft]

360 should be good to go and might be useful here as well. Hopefully not too many conflicts 😬

[HChughtai]

Looking into why the whitelisting may be causing an issue with DICOM to Nifti conversion, it does seem like this is due to too aggressive anonymisation removing needed MRI image tags.

It's possible that this is due to required (type 1) attributes in the MR image module being removed. See also https://dicom.innolitics.com/ciods/mr-image/mr-image for a more readable format.

[HChughtai]

Used https://github.com/pydicom/dicom-validator to look at the intermediate anonymised DICOMs, and there were a few things that are being flagged up as missing from a fully valid MR compared to its IOD. The minimum needed to be added are:

• (0020,0032): "Image Position (Patient)"
• (0020,0037): "Image Orientation (Patient)"
• (0018,0023): "MR Acquisition Type"

[peshence]

Closing in favour of #366