refactor(stack): drop Helm-ownership migration hacks

[OisinKyne]

The 9ef8cda commit added two pieces of compatibility ballast that
ran on every obol stack up:

1. migrateBaseHelmOwnership — labelled and annotated a hardcoded list
   of pre-existing namespaces/RBAC objects to hand them to Helm.
   The trigger was a one-off transition where prior obol versions
   created those objects outside Helm; brittle (any new resource
   needs another entry) and only exists to paper over a known dev-
   machine state.

2. needsLiteLLMConfigHelmMigration / the conditional delete-and-
   restore branch in preserveLiteLLMConfigForHelm — fired when the
   litellm-config ConfigMap had a non-helm field manager
   (typically kubectl-patch from old obol model setup calls).
   Current model.go patches use --field-manager=helm explicitly,
   so the trigger is unreachable in the current code path.

Both fall under "code for a one-off issue that already passed".

What stays: preserveLiteLLMConfigForHelm (simplified to always
snapshot), restoreLiteLLMConfig, mergeLiteLLMConfig, and
ApplyServerSideForceConflicts. The base helmfile template only
includes the paid/* catch-all route; without preserve+restore,
every obol stack up would wipe user-added cloud providers and
custom endpoints. That's a real-resilience exception, not a hack.

Tests: TestNeedsLiteLLMConfigHelmMigration removed.
TestMergeLiteLLMConfigPreservesChartDefaultsAndPreviousModels stays.

Co-Authored-By: Claude Opus 4.7 (1M context) noreply@anthropic.com

[bussyjd]

Superseded by #386 — all commits from this branch are already present at the tip of integration/pr377-pr381 (verified by git log <branch> ^origin/integration/pr377-pr381 returning empty). Closing to keep the queue tidy. Track final landing on main via #386.