aggregations: multi-event, multi-field aggregator

[slint]

Design an Aggregator class that can aggregate different kind of events and fields into a single document. This satisfies the need to keep track of statistics on e.g. the record level (record views, record downloads, etc).

Example produced document:

# Index "stats-record-2018-06"
{
    '_id': "recid-1234-2018-06-01",  # one document for each recid and day
    'recid': '1234',
    'conceptrecid': 'C_1234',  # included for aggregated version queries
    'doi': '10.5555/1234',
    'conceptdoi': '10.5555/C_1234',
    
    # aggregated from "record-view" events
    'views': 100,
    'unique_views': 60,
    'machine_views': 100,
    'unique_machine_views': 60,
    
    # aggregated from counting "file-download" events
    'downloads': 100,
    'unique_downloads': 60,
    'machine_downloads': 100,
    'unique_machine_downloads': 60,
    # aggregated from summing up the "size" field from "file-download" events
    'download_volume': 100,
    'unique_download_volume': 60,
    'machine_download_volume': 100,
    'unique_machine_download_volume': 60,
}

Example config for the aggregation:

dict(
    aggregation_name='record-agg',
    templates='invenio_stats.contrib.aggregations.aggr_file_download',
    aggregator_class=StatAggregator,
    aggregator_config=dict(
        client=current_search_client,
        doc_id_field='recid',
        aggregations={
            'views': {
                'event': 'record-view',
                'field': 'recid',
                'filters': ['no_robots'],
            },
            'uq_views': {
                'event': 'record-view',
                'field': 'recid_session_id',
                'filters': ['no_robots'],
            },
            'machine_views': {
                'event': 'record-view',
                'field': 'recid',
                'filters': ['no_robots', 'is_machine'],
            },
            'uq_machine_views': {
                'event': 'record-view',
                'field': 'recid_session_id',
                'filters': ['no_robots', 'is_machine'],
            },
            'downloads': {
                'event': 'file-download',
                'field': 'recid',
                'metric': ('cardinality', 'recid + visitor_id'),
                'filters': ['no_robots'],
            },
            'uq_downloads': {
                'event': 'file-download',
                'field': 'recid',
                'metric': ('cardinality', 'recid + session_id'),
                'filters': ['no_robots'],
            },
            'download_volume': {
                'event': 'file-download',
                'metric': ('sum', 'size'),
                'field': 'recid',
                'filters': ['no_robots'],
            },
            'uq_download_volume': {
                'event': 'file-download',
                'metric': ('sum', 'size'),
                'field': 'recid + session_id',
                'filters': ['no_robots'],
            },
        },
        aggregation_interval='day',
        copy_fields=dict(
            file_key='file_key',
            bucket_id='bucket_id',
            file_id='file_id',
        )
)))