Allow more customization of datasource (e.g. connectionInitSql)

[strangemonad-faire]

There are many reasons for wanting to set (possibly driver-specific) options on the datasource config. For example, we need to configure the client connection to properly use utf8mb4 everywhere (it's not purely a server-side config). For this to work, we'd ideally set the characterSetEncoding and characterSetResults and connectionCollation jdbc url query params in addition to being able to override the hikari connectionInitSql property to "SET NAMES [charset] COLLATE [collation].

There are a couple options I could think of, they aren't necessarily mutually exclusive but their interactions might have to be documented.

1. Allow users to provide jdbc url params (that would get appended to the ones build in #buildJdbcUrl) with the documented caveat that the params might be driver specific.

2. Given that hikari config has been battle tested and covers a wide range of needs at this point, either expose HikariConfig directly or make DatasetConfig fields a superset of the hikari config options.

[swankjesse]

In the spirit of offering opinionated defaults, I'd like both better configuration options and better defaults.

[strangemonad-faire]

Here's a concrete proposal:

1. Better defaults:

I'd love if misk-hibernate had done the "right thing" here. We're in a unicode / utf8 world. Mysql messed up with the whole utf8 vs utf8mb4 and most people don't care to know the details. Had the misk-hibernate default been set up for utf8mb4 for new services, we wouldn't have gotten into this situation needing to migrate some existing tables. Additionally, mysql > 5.7 mostly fixes the whole VARCHAR(191) limit for indexable columns. It gets a bit complicated because some of these settings either need to be set on the server or or come about from an interaction of server and client settings

That said, I'd love if, similar to how migration versions are checked on startup, expected db settings are checked. For 5.7-innodb that would be something like:

• create table default engine innodb
• create table default for format DYNAMIC (https://dev.mysql.com/doc/refman/8.0/en/innodb-row-format.html)
• make sure innodb is set for large prefix index (to get around the whole varchar(191) https://dev.mysql.com/doc/refman/5.6/en/innodb-parameters.html#sysvar_innodb_large_prefix (this setting is going away, large prefix will be the default in later versions.
• The 00:00 timezone is already taken care of
• Row format DYNAMIC (https://dev.mysql.com/doc/refman/5.6/en/innodb-row-format.html) allows large prefix index
• Set names utf8mb4 collate utf8_unicode_ci
• Check that all the appropriate charset and collation variables are correctly setup for utf8mb4.

The idea being that if you want to connect to a pre-existing datasource that is not setup for unicode or is in transition, you have to specify that in your misk config, the default gives you the right behaviour for new DBs.

2. Better config options:

The landscape of connectors, drivers, database engines, versions etc is large. I think there needs to be some kind of abstraction escape hatch. Expecting misk to cover all this api surface area would be pretty gnarly, there's always going to be something missing. E.g. we're running on mysql-aurora, while the mysql connectorJ driver works ok, the recommended driver is the miradb "smart" connector that knowns how to handle failover.

[strangemonad-faire]

also, hi @swankjesse :)