Fixed
* [BUG]: used `ahash_1` instead of `ahash_2` variable
Added
* draft installation steps for restricted windows
Changed
* [hasher]: make it faster using mp + introduced dependency on `dill` package
* [hasher]: implement the `missing values` rule for RACE and SEX if value in ['NI', 'UN', 'OT'] ==> no hash is generated
* [linker]: create a UUID even for patients with zero hashes
* Allow nulls in the `linkage_hash` column since some patients do not have all required data elements
* [schema]: update linkage table with `rule_id` + `fk_linkage_rule_id`
* [linker]: fix error with un-set flag + do not create tables using SQLAlchemy
* [linker]: update example config file file
* logging: search the config file under `cwd`
* Split requirements into two to reduce dependencies for pypi installs
* [linker]: add logic for logging hashes that resolve to two distinct UUIDs
* [linker]: update schema definition for UUID use varchar(32) instead of binary(16)
* [linker]: changed the model to use a `text` field for UUIDs instead of `binary` (for compatibility with the PCORI tables)
* [linker]: when there is only one hash provided store the link object in the proper slot
+ return a link when hashes resolve to two different UUID's
* [linker]: implement 2 changes per Jiang - if only one of the two hashes is found re-use existing UUID
(previously we were giving precedence to the `rule_1` hash by generating a new UUID)