[Draft] Initial Redshift Driver Implementation

[eliasdefaria]

Putting up a draft for what I was able to get done so far at the Seattle on-site!

Supported Types

(Redshift Type -> Arrow Type)
int2 -> int16
int/int4 -> Int32
int8 -> Int64
float4 -> float32
float/float8 -> float64
bool -> boolean
char/varchar/text -> string
date -> date32
time/timetz -> time64us
timestamp/timestamptz -> timestamp_us
decimal/numeric -> decimal128

Things I would like to do but didn't get a chance to:

• Additional testing to make sure all the data types are handled correctly
• More robust casting of data types in the record reader when they're passed to the builders
• Testing
• Adjusting credentials handling to be easier to understand and set. This part is a bit of a challenge to get right since different auth methods require different inputs and therefore options.
• Optimizing performance of the record reader. Would like to also clean up the code a bit here as part of that effort. Because the schema comes with the first batch of records, we can do a lock-step sort of synchronization that Felippe implemented really nicely for us in Databricks

Some interesting tidbits that were discovered:

• Redshift Data API requires auth to AWS via IAM credentials (https://docs.aws.amazon.com/redshift/latest/mgmt/data-api-iam.html)
• Username password over the wire without IAM isn't supported here
• Authenticated calls also require authentication to the database via IAM identity center (not supported in this PR), or AWS SM stored username/password for a DB user, or a temporary user (provide a username). (https://docs.aws.amazon.com/redshift/latest/mgmt/data-api.html#data-api-calling-considerations-authentication)
• Internal conditions dictate the size of the pagination behavior

Cheers,
-Jason :)