Firepit

Minor `query` fixes:
- tweak to `Column` validation
- use `Column` internally in `Predicate`

Bug fix for PostgreSQL - fix errors like this:
DependentObjectsStillExist: cannot drop view conns because other objects depend on it
DETAIL: view ts_conns depends on view conns
HINT: Use DROP ... CASCADE to drop the dependent objects too.

- added `splint convert`: CLI to convert some logs to STIX 2.0 bundles of `observed-data` SDOs. Supported formats include:
- Windows datasets from https://securitydatasets.com
- some Zeek logs (conn.log and dns.log should work)
- "flat" JSON (each object MUST contains `type` property and the required STIX properties for that type)
- bugfix for postgres: fix view def rewriting to handle double-quoted column names
- this would cause `UndefinedColumn` exceptions on attempts to use new columns added e.g. by `reassign` (i.e. Kestrel `APPLY` commands)

Bug fixes:
- add explicit `close()` API
- postgres: map `UndefinedColumn` to `InvalidAttr` in `_create_view()`
- postgres: fix `_get_view_def()` to handle `UNION`s
- query: don't assume `rhs` arg in `Predicate` is a string

Fix an issue in the reassign function that only happens on postgres:

File "/opt/modules/lib/python3.9/site-packages/firepit/sqlstorage.py", line 475, in reassign
splitter.write(obj)
File "/opt/modules/lib/python3.9/site-packages/firepit/splitter.py", line 243, in write
self.writer.new_property(obj_type, col, col_type)
File "/opt/modules/lib/python3.9/site-packages/firepit/splitter.py", line 109, in new_property
self.store._add_column(tablename, prop_name, prop_type)
File "/opt/modules/lib/python3.9/site-packages/firepit/pgstorage.py", line 277, in _add_column
self._execute(f'DROP VIEW IF EXISTS "{viewname}"', cursor)
File "/opt/modules/lib/python3.9/site-packages/firepit/sqlstorage.py", line 146, in _execute
cursor.execute(statement)
File "/opt/modules/lib/python3.9/site-packages/psycopg2/extras.py", line 236, in execute
return super().execute(query, vars)
psycopg2.errors.DependentObjectsStillExist: cannot drop view process_0 because other objects depend on it

Minor fixes plus some new internal APIs:
- add a `set_table` method to query Filter and Predicate in order to retroactively disambiguate columns specified in query. Only needed in some special cases, like query generation in Kestrel.
- add a `paths` keyword arg to `auto_deref` for more fine-grained control of which reference properties get auto-derefenced.