Continuous graph population - Ether Graph tasks

[shashankpr]

To track the tasks pending before continuous graph population can be enabled in production are listed below:

• Standardize graph schema and connections through python structs
• Add the following connections:
• User --> Entities, User --> User (groupedWith), User --> User (recommended/related), Entities --> Keyphrase (relatedTo)
• Populate only summary phrases (Populate and persist only summary-segments #163)
• Add logic to cleanup graph after context.instance.ended and remove non-highlight phrases.
• Fix dateTime format issue in production data and set dateTime as index for sorting (Dgraph index type changes #156)
• Add support for Facets (Add Facets to Graph schema #182)
• Separate Dgraph client as a common pkg in ai-engine or api

[shashankpr]

• Standardize graph schema and connections through python structs
• Add the following connections:
• User --> Entities, User --> User (groupedWith), User --> User (recommended/related), Entities --> Keyphrase (relatedTo)
• Populate only summary phrases (Populate and persist only summary-segments #163)

The above tasks have been implemented in #170 and explained in #163

[shashankpr]

Add logic to cleanup graph after context.instance.ended and remove non-highlight phrases.

Tagging segments as highlight=True to query high-fidelity data. The non-highlight segments are not deleted as they could be useful for other tasks

Fix dateTime format issue in production data and set dateTime as index for sorting (#156)

PR #188 addresses this issue. Also, ContextSession node has a new predicate - startTime that denotes the meeting's start time. This predicate is indexed for sorting and querying so that we can get latest N meetings deterministically.