Add repo connectivty checks to verify (preflight) playbooks

[kvelarde-itential]

Made verify.yml continue past a single component's failure instead of aborting the whole run.
Root cause: Ansible aborts the entire ansible-playbook run — not just the current play — the moment a play ends with 100% of its hosts failed, even for later, unrelated plays. Since verify.yml chains verify_redis/verify_mongodb/verify_platform/verify_gateway via import_playbook, one bad host in Redis could prevent MongoDB/Platform/Gateway from ever being checked.

• Made the OS/architecture asserts in verify-host.yml non-fatal (ignore_errors + register: os_validation/arch_validation), matching the existing CPU/RAM/disk pattern.
• Made the final combined assert in verify-results.yml non-fatal, with every registered result wrapped in | default({'failed': false}) so a component that skips a check (e.g. verify-sentinel.yml never calls verify-connectivity.yml) doesn't crash the assert on an undefined variable.
• Added one real, un-ignored final play to verify.yml ("Report Overall Verification Result", hosts: all) that asserts on a per-host verification_passed fact — since nothing runs after it, failing there is safe, and it's what gives the whole run a genuine non-zero exit code while still letting every component's checks run against every host.

Fixed a cross-component overwrite bug. verification_passed (and later validation_errors) were single shared per-host facts that got clobbered by whichever component ran last. On a host belonging to more than one component group (e.g. an all-in-one host), a later component passing cleanly erased an earlier component's failure on the same host. Fixed both to accumulate (logical AND / dict merge via combine()) instead of overwrite.

Added per-component failure reporting. The final report play now shows a component_validation_errors dict keyed by component ({"Redis": [...], "Platform": [...]}) instead of just a pass/fail flag, so it's immediately clear which component failed which specific check on a given host.

Removed the Vault connectivity pre-check. The standalone Vault-installing role was dropped from this collection in v4 — Platform now only configures a client connection to an externally managed Vault, so checking access to Vault's install RPM repo (rpm.releases.hashicorp.com) pre-install was stale. Removed from common_required_repositories, verify-platform.yml, and the README repository table. Also renamed the copy-pasted "Verify Platform host requirements" task name in each component's results-reporting step to match the actual component (Redis/MongoDB/Platform/Redis Sentinel).

Fixed a Redis Sentinel connectivity bug. verify-sentinel.yml's connectivity check used component_name: "Redis Sentinel", which matched no entries in the repository list — it silently verified nothing. Fixed to check Redis's own repositories, since Sentinel-only hosts still install the full Redis package/source (same repos are needed).

Moved required-repository lists out of common and into each component role. common_required_repositories was a single shared list in roles/common/vars/main.yml, filtered by component_name at verify time. Split into redis_required_repositories, mongodb_required_repositories, platform_required_repositories, and gateway_required_repositories, each now owned by its own role's vars/main.yml (created one for Gateway, which didn't have one). verify-connectivity.yml's interface changed accordingly — it now takes the list directly via required_repositories instead of filtering a shared list by name. roles/common/vars/ no longer exists.

Added Gateway hardware-spec verification (built on top of the gateway_hw_specs var and common:verify-host wiring). Found and fixed a bug where the new CPU/RAM/disk/OS/arch checks were being collected into validation_errors but never actually gated in the final assert (which only checked connectivity_validation) — a Gateway host failing hardware requirements would still report success. Switched Gateway to the same verify-host → verify-connectivity → verify-results sequence used by Redis/MongoDB/Platform, so it's now consistent and hw-spec failures actually fail the check.

Docs: updated roles/common/CLAUDE.md, roles/redis/CLAUDE.md, roles/mongodb/CLAUDE.md, roles/platform/CLAUDE.md, roles/gateway/CLAUDE.md, top-level CLAUDE.md, and README.md throughout to match.