Fix an issue where summary data would be interpolated

[eivindjahren]

This removes the following unwanted behaviors of the current implementation:

• The call to _init_numpy_vector_interp changes some values, probably due to interpolation.
• Iterating over ecl_sum uses both ecl_smspec_node->key1 and ecl_smspec_node->key2 which will result in duplicate entries for summary keyword types that have both types.
• DATE is included in the values loaded

However, the following behavior is kept from the most recent implementation:

• All report steps are loaded, not just those in the time map.

Pre review checklist

• Read through the code changes carefully after finishing work
• Make sure tests pass locally (after every commit!)
• Wait for tests in CI to pass (see "checks" below)
• Prepare changes in small commits for more convenient review (optional)
• PR title captures the intent of the changes, and is fitting for release notes.
• Updated documentation
• Ensured new behaviour is tested (opened GUI? screenshots? tried workflows?)

Pre merge checklist

• Added appropriate release note label
• Commit history is consistent and clean, in line with the contribution
  guidelines.