Kart should check PK fields are actually unique

[craigds]

Describe the bug
If you import a tabular dataset with the --primary-key option, it's possible to use a field that isn't actually unique.

Kart doesn't check, and any duplicate rows silently overwrite each other during import.

To Reproduce

• sqlite3 bad-pks.sqlite

• create table mydataset (badpk integer, sometext text);
  insert into mydataset (badpk, sometext) values (1, 'text1');
  insert into mydataset (badpk, sometext) values (1, 'text2');
  

• kart init myrepo ; cd myrepo
• kart import --primary-key=badpk OGR:../bad-pks.sqlite
• kart show -- mydataset:feature

Expected behaviour

An error should be raised since the badpk field has two duplicate values

Output

The dataset is imported and only one of the features survives

$ kart show -- mydataset:feature
commit d2439097269d5dccd73aaec4d0fd388e6e9aa669
Author: Craig de Stigter <craig@destigter.nz>
Date:   Fri Jun 23 10:21:11 2023 +1200

    Import from bad-pks.sqlite:mydataset to mydataset/

+++ mydataset:feature:1
+                                    badpk = 1
+                                 sometext = text2

**Version Info **

• OS: macOS 13.4
• Version: 0.14.0rc2

[hamishcampbell]

Duplicate of #307 ?