Add more guardrails around data type expectations

[firste]

Is your feature request related to a problem? Please describe

#92 adds stricter alignment between the data dictionary and the data in seismometer. This brings with it some assumptions around data typing related to Pandas support for nullability and sklearn support for bools that may not align with expectations.

Describe the solution you'd like

Add more guardrails around data type expectations. A few ideas -

• show a warning when target columns are bool
• update the data dictionary creation code mentioned in Readd data dictionary creation features #93 to be aware of these assumptions.