Mixed downloads

[alycejenni]

The primary change for this PR is the addition of the "mixed" category of downloads (alongside "collections", "research" and "gbif"), which is to account for downloaded files that contain records from both collections and research datasets. This only affects download_events.

For example:

• resource_1 is a collections resource
• resource_2 is a research resource
• download_1 contains 100 records from resource_1
• download_2 contains 100 records from resource_2
• download_3 contains 100 records from resource_1 and 100 records from resource_2
• so the total events is 3 (2 collections and 2 research), total records is 400 (200 collections and 200 research)

Previous this would have been represented as (ignoring the "gbif" category):

{
    "collections": {
        "download_events": 2,  # download_1 and download_3
        "records": 200  # download_1 and download_3
    },
    "research": {
        "download_events": 2,  # download_2 and download_3
        "records": 200  # download_2 and download_3
    }
}

There's no way to establish that the total number of download events of any type is 3,

This is now represented as (again ignoring the "gbif" category, which is unchanged):

{
    "collections": {
        "download_events": 1,  # download_1
        "records": 200  # download_1 and download_3
    },
    "research": {
        "download_events": 1,  # download_2
        "records": 200  # download_2 and download_3
    },
    "mixed": {
        "download_events": 1,  # download_3
        "records": 0  # always 0
    }
}

Representing it like this means that the totals are accurate, even if the representation is initially somewhat confusing. From this, we can calculate that the total number of all download events is 3, the total number of download events involving a collections dataset is 2, and the total number of download events involving a research dataset is 2.