Support JSON extract path expression

[heywhy]

Previously, queries that check against JSON columns couldn't be used alongside this adapter. With the changes in this PR, we can now do this. see an example below:

from(e in Model.Employee)
|> where([e], e.metadata["twitter"] == "@andrew_fuller")
|> Repo.one!()

Also, a function called flush_tables was added to the adapter module to allow us to drop all data in a repository considering some tests might pollute the database and we have no way to implement postgres-like partitioning at the moment.

[evadne]

@heywhy Thank you very much, I am reviewing this now.

[evadne]

My initial view: Very pleasant set of changes, good job and thank you for the contribution. Please can you review my comments and advise. Thank you again for your time.

[evadne]

@heywhy FYI, reconstructing branch as your changes were from 0.1.6 and included undesired cross merge from master 1.0.1 so history needs to be refactored.

[evadne]

Reconstitution of WIP @ feature/gh-20-json-extract-path

[evadne]

@heywhy Please retest branch feature/gh-20-json-extract-path @ 31f2c3c, then pull this to your fork (fast forward only, do not create merge commits), then revise the working branch in this PR to allow correct attribution via GitHub history.

Please note active_tables is removed, you should do the following instead,

for model <- [ModelA, ModelB] do
  Repo.delete_all(model)
end

As I have implemented delete_all support.

[evadne]

@heywhy do not push to develop, use the same branch I published and revise this PR to use that branch to conform to merge requirements.

[heywhy]

@heywhy do not push to develop, use the same branch I published and revise this PR to use that branch to conform to merge requirements.

That's a mistake on my end. I thought you wanted the commits squashed or something. I will close this PR and create a new PR from the other branch as everything from the other branch works as expected.

[evadne]

Yes that would be good though name the branch with feature/gh-21 or so. @heywhy

[heywhy]

Also, can you have a part in the readme that tells contributors what merge strategy should be used so as to avoid a situation like mine?