DM-41029: Add metadata table to APDB schema

[andy-slac]

This also adds version number to APDB schema and "advances" it to 0.1.1, all schema versions before metadata are considered 0.1.0.

[gpdf]

@andy-slac Can you describe how the new metadata table will be used?

[andy-slac]

Can you describe how the new metadata table will be used?

One purpose is to hold schema version, some details are in https://dmtn-269.lsst.io/. There may be other uses to store per-database information, e.g. instrument name. This should be considered as implementation detail, I'm not even sure that it needs to be in sdm_schemas, but my impression is that people prefer to have it here.

[cbanek]

Seems reasonable

[parejkoj]

Just to confirm: this would make the stored version numbers strings, so we'd have to do some version parsing with an appropriate library to compare them? I think that's fine, I just wanted to check.

[parejkoj]

Should we call this key, in keeping with python dictionaries (which this effectively is), or is name the typical database term for this kind of key:value store?

[andy-slac]

key is more natural, but in some database systems key is a reserved keyword and needs special care when used as column name (proper quoting). It's easier to avoid using keywords for column names.

[andy-slac]

Just to confirm: this would make the stored version numbers strings, so we'd have to do some version parsing with an appropriate library to compare them? I think that's fine, I just wanted to check.

Right, all values are stored as strings in the metadata table and need de/serialization if original type is different from string. This will need client-side code to do that transformation. It's hard to implement generic type-safe storage, at this point I think we are OK with handling this entirely on client side. Version numbers for APDB will be stored in SemVer format (X.Y.Z) and there is a class to support that format, for other potential cases I think we can introduce some helper methods when necessary.

[gpdf]

@andy-slac We've been fiddling with some of the test workflows in this repo. You may want to trigger a re-run of the tests.