[reporting] Per-tenant mapping of "searchable" attributes

To circumvent the limitations imposed by Elasticsearch having a maximum of 1000 unique attribute key names, we will use MongoDB to provide a persistent per-tenant attribute enumeration mapping. The mapping needs to be symmetric, mapping keys to integers for encoding and from integers to keys for decoding attributes. Moreover, updates to the maps must be atomic. A new searchable attribute should be added for each new attribute key a device submits until the tenant has exceeded 100 attribute keys.