Impove performance of map parsing routines.

[ForeverYoung1208]

Now it takes several minutes to load and parse map with words quantity around 100. It looks like word node creation methods not quite optimized for creation of several words, so we need to investigate and optimize these routines to make map parsing quicker.

[ForeverYoung1208]

I tried to improve performance of loading maps. (Which is mainly depend on creating word nodes with all necessary properties)

I tried several approaches, such as

• making special lightweight methods at graph services (without checking on existance of the node to be created)
• inserting all nodes to be created in one SQL statement (and then add necessary properties to each node)
• replacing some methods from typerom querybuilder to raw sql insert queries (but note that these sqls are executed using typeorm's query runner anyway, but without accessing entities). But, even if i've reached any perforemance improvement, it is barely visible.

Also I noted that there no big diference between typeorm inserts and raw sql inserts, so i think for future performance analysis it is worth to look at more deep restructuring of praph services, and try to reduce the amount of queries for high - loaded operations, i.e. make sql queries to insert all property_keys and all property_values at once, if it is possible.

Maybe, even try to find some replacement of the stack sql.js + localforage + indexed db (if such a replacement exists) But it will require additional research.