Add generic testscript commands for testing against
StateDB tables.

This allows implementing tests as scripts, which becomes useful when
tests perform multiple steps on tables and need to verify the output
each step.

Signed-off-by: Jussi Maki <jussi.maki@isovalent.com>

Add a test using testing/quick for StateDB to validate queries against
unique and non-unique indexes.

This catches the two issues with non-unique indexes:
- Prefix and LowerBound did not properly seek according to the search key
  (e.g. secondary key) and matched against the composite instead, e.g.
  if we have composite keys <a><aa><1> ("aaa1") and <aa><b><2> (aab2)
  and prefix search for "aa" we'll yield the object with composite key
  <a><aa><1> incorrectly as we did not check if the secondary key length
  (1) matches with search key "aa" (2).

  (non-unique keys are encoded as <secondary><primary><secondary length>,
   see encodeNonUniqueKey for details)

- Since there's no separator between the secondary and primary key in the
  encoding, the iteration order for <a><z><1> and <aa><a><2> will be
  <aa><a><2>, <a><z><1>, but it should be <a><z><1>, <aa><a><2> since
  "a" < "aa".

Signed-off-by: Jussi Maki <jussi.maki@isovalent.com>

Add a regression test to validate searching non-unique indexes.
This more clearly shows the issues found by the TestDB_Quick that
was added by the previous commit. See the previous commit for more
details.

Signed-off-by: Jussi Maki <jussi.maki@isovalent.com>

Prefix and LowerBound searches did not properly deal with non-unique
indexes. In these indexes the keys are encoded as <secondary><primary><secondary len>,
and prefix searching needs to make sure the secondary key length on the results
is equal or longer than the search key as otherwise we might match into the primary
key.

For example if the object is struct{A, B string}, with A being primary and we have
{"a", "a"}, then the secondary index is key'd as "aa<1>". A prefix search with "aa"
must not match since the secondary index contains only an object with key "a".

Fix this by using a special iteration on non-unique indexes that checks the length
of the secondary key and ignores any other matches that are due to matching into the
primary key.

Another issue with non-unique indexes was due to having no separator between the
secondary and primary key, leading to the primary key having an effect on the iteration
order. Fix this by adding '\0' as a separator and base64+sha256'ing the primary key.

Signed-off-by: Jussi Maki <jussi.maki@isovalent.com>

Bring in the new hive version that includes the script support.
Fix the API usage.

Signed-off-by: Jussi Maki <jussi.maki@isovalent.com>

To allow using part.Set[] and part.Map[] types with the YAML marshalling/unmarshalling
in the StateDB script commands, add the custom YAML marshalling and unmarshalling methods
for these types.

Signed-off-by: Jussi Maki <jussi.maki@isovalent.com>

Signed-off-by: Jussi Maki <jussi.maki@isovalent.com>

Reimplement the reconciler tests using scripttest. This significantly
simplifies the test-suite and allows easier verification of more complex
scenarios.

To allow for Status JSON and YAML marshalling, define custom UnmarshalJSON
and UnmarshalYAML that also fill in the 'id'. The 'id' is used with StatusSet
to efficiently allow multiple reconcilers to manipulate the status without
conflicts, e.g. if the object status id is the same it can still be updated
with the reconciliation result even if the object conflicted due to other
reconciler's update to its status.

Signed-off-by: Jussi Maki <jussi.maki@isovalent.com>

Add *String counterparts (Int32 => Int32String) for implementing FromString
in Index[].

Signed-off-by: Jussi Maki <jussi.maki@isovalent.com>