Quick fix to serialize load metrics as list instead of a dictionary

[sultaniman]

This issue was reported by a community member in slack and related issue #1043

When we capture load_info data in the destination database the following occurs:

• Pipeline execution slows down significantly after x number of incremental pipeline runs (it is back to the original speed when load_info is not captured in the database). My specific job (sourcing data from rest API capturing it in MotherDuck) slows down from 1.5 minutes to over 10 minutes (after a few hundred runs) and seems to keep getting slower.
• Each pipeline run creates a new table with one record (at least in my simple pipeline) with names like _load_info__metrics___1709356777_4431682.

TODO

• Create 2 separate sources to load
  1. One has resource w/ primary key
  2. Another with two resources
  3. load both sources
• load the first source
• load extract, normalize and load info
• run pipeline with schema parameter

[netlify]

✅ Deploy Preview for dlt-hub-docs canceled.

Name	Link
🔨 Latest commit	4839541
🔍 Latest deploy log	https://app.netlify.com/sites/dlt-hub-docs/deploys/65e772b45d08cc00086460d4

[rudolfix]

1. We do not need another test file. Please add it to pipelines tests.
2. we do not use faker (yet). please revert lock and pyproject for now
3. why those primary keys are added to run? you have them in the resources

[rudolfix]

LGTM now,

@sultaniman

1. there were bugs in test (loading tables into trace schema)
2. there was an inconsistent way to replace schema content (bug got fixed)