[API tests] Separate APIV1 fixture and verification reliability fixes

[Prangshuman Das (t-prda)]

Scope

AB#646383

Draft follow-up to #11230, based on prdas/646383-split-return-shipment-pdf.

• Identify five posted blank API lines by GUID description and exact record lookup.
• Regenerate RapidStart fixtures per invocation; refresh all three polling loops, correct expected/actual assertion order, and fix the apply-status qualifier.
• Keep G/L journal-template handler state in a codeunit global.
• Enable Foundation setup for the invoice discount-percent scenario.
• Re-enable exactly ten APIV1 methods temporarily quarantined while these business fixes were separated. These were conservative/deferred exclusions, not newly proven failures. RS create/upload were not additionally quarantined.

Lifetime and preservation

Automation initialization keeps workdate first, followed by idempotent authentication-provider selection, per-test cleanup, and sample generation. Removing the guard makes provider selection per invocation; its existing same-provider early return remains intact.

Preserves all seven earlier layers, APIV2 wildcards, and all 22 Expense exclusion entries. Only eight AL files and the APIV1 exclusion manifest change.

Validation

• Static context-patch and reviewed-method equivalence checks passed.
• Exact nine-file scope, protected-manifest byte equality, and git diff --check passed.
• This stacked head remains runtime untested. Prerequisite [Expense Agent tests] Fix action-result assertions for runtime compilation #11340 has resolved the CU 148339 compilation blocker; that success does not validate this layer. The RapidStart polling tests require background task execution; the parent now explicitly configures this for the existing UncategorizedTests profile. That profile still needs fresh CI verification, and it does not validate the other business fixes in this layer. No local NST was used.

Linked to native GitHub stack #11232: #10085 is the bottom auth PR, followed by #11224 through #11230, then this draft as the top layer. All eight follow-ups are drafts. No branch was force-pushed and no PR was merged. The first seven layers restore 37 deferred methods; this layer restores the remaining ten, preserving pre-existing exclusions.