Cache ClassInfo methods results

[nithril]

Expected Behavior

Per my (little) understanding of the ClassInfo class I would suggest to cache the results of the following methods:

• ClassInfo#relationshipFields
• ClassInfo#getEndNodeReader
• ClassInfo#startNodeReader

Context

On a project I'm working on we have performance issues. We made the exercise to profile the application to fix the bottleneck.
We have identified the aforementionned methods as responsible of a fraction of the time taken (not negligible).

Please let me know if it makes sense. If so I can submit a pull request.

Your Environment

• Java version: 1.8.0.265
• Neo4j version (The version of the database): 3.5.21
• OGM Version (The version of this software): 3.2.17
• OGM Transport used (One of Bolt, HTTP or embedded): Bolt
• Neo4j Java Driver version (in case of Bolt transport): 4.0.2

[meistermeier]

Thanks for reporting this (and the two others).
I just had a glance at the methods you mentioned and think that a storage of those static information could be worth the memory.
So if you are willing to submit a PR for those, go ahead. Looking forward to see your contribution.
Even though this one and #853 are pretty close together, I would prefer to have two separated PRs to avoid too much changes in a commit.

[nithril]

👍 I will work on that. Given that #849 has already tackled ClassInfo#relationshipFields I will do the 2 others one