Support complex data type on node properties

[QubitPi]

[!NOTE]

This is a followup of https://github.com/paion-data/aristotle/issues/23

One of the reasons why many people use yahoo Elide is its ability to support arbitrarily complex entity graph in a single entity attribute. This useful feature maps successfully between business domain data and its database representation in SQL.

This is something Aristotle will do as well, because our application is now facing the same issue of mapping the business data into the database representation.

On nearly half of our nodes, there is a "table property". For example, suppose we have a node with the following attributes (I'm presenting it in YAML):

  - term: der Gegenstand
    definition:
      - object
      - thing
    declension:
      - ["",         singular,                    plural      ]
      - [nominative, Gegenstand,                  Gegenstände ]
      - [genitive,   "Gegenstandes, Gegenstands", Gegenstände ]
      - [dative,     Gegenstand,                  Gegenständen]
      - [accusative, Gegenstand,                  Gegenstände ]

[!TIP]

The data above become a sub-graph of 3 nodes, we can see what our production data looks like by following these steps

1. Go to "Viewing Our Production Data in Docker" and follow the steps there
2. Execute this query: MATCH (term:Term)-[r]-(x) WHERE term.name = "der Gegenstand" and term.language = "German" RETURN term, r, x
3. We should be able to see something like this:

2 attributes (term, declension) above represents 2 node properties. The declension table is the same thing as the table below

	singular	plural
nominative	Gegenstand	Gegenstände
genitive	Gegenstandes, Gegenstands	Gegenstände
dative	Gegenstand	Gegenständen
accusative	Gegenstand	Gegenstände

Knowing that Neo4J does not support object property, we flatten out each table cell, each one of which is a Neo4J node property. For example,

term: "der Gegenstand" declension-0-0: "" declension-0-1: "singular" declension-0-2: "plural" declension-1-0: "nominative" declension-1-1: "Gegenstand" declension-1-2: "Gegenstände" declension-2-0: "genitive" declension-2-1: "Gegenstandes, Gegenstands" declension-2-2: "Gegenstände" declension-3-0: "dative" declension-3-1: "Gegenstand" declension-4-2: "Gegenständen" declension-4-0: "accusative" declension-4-1: "Gegenstand" declension-4-2: "Gegenstände"

This makes the decoding at frontend pretty wacky. Considering we might have 2- or 3-level nested structure in the future, this approach needs to be improved sooner rather than later.

With this issue, the Aristotle is facing an opportunity to become a more useful product than just being a database proxy - mapping business domain model to database representation and vice versa. We should make Aristotle support complex data on node property from the frontend perspective.