fix(installer): memory check flip-flops between WSL VM budget and physical RAM; impossible recommendations

[LukasWodka]

Why

Get-PfMemGb prefers docker info MemTotal over physical RAM, so the same 15 GB laptop reports "Memory: 7 GB" when Docker is running (WSL2 VM default = 50% of RAM) and "Memory: 15 GB ✓" when it isn't — flip-flopping across re-runs and confusing users mid-install. Worse, the remediation hint told a 15 GB machine to "give Docker >= 16 GB" — physically impossible advice. Observed on three separate machines (7 GB, 6 GB, 15 GB readings; the 16 GB hint on a 15 GB host).

Where

scripts/install-k8s.ps1 (v1.9.6): Get-PfMemGb/Get-PfRuntimeMemGb L2256–2277, hints L2351/L2356/L2425, Step-2 re-check Test-PreflightRuntimeMem L1035.

Scope

• Always print two labeled lines: host RAM and Docker's current budget (when a runtime is up).
• Cap every recommendation at (physical − 2 GB); never recommend more than the machine has.
• De-duplicate the Step-1/Step-2 warnings into one consistent message.

Acceptance criteria

• Same machine, Docker up or down, reports identical host RAM; recommendations are always achievable.

[LukasWodka]

Cross-OS audit result: same bug on macOS; Linux is benign (bash v1.9.6 evidence):

• Same prefer-the-runtime pattern: scripts/lib/preflight.sh L111–112 (docker info --format '{{.MemTotal}}') with host fallback (sysctl -n hw.memsize, L126); the printed label flips between "host" and "Docker VM" depending on whether Docker is up (L227, L235) — cold run says 16 GB, warm re-run says 6 GB on the same Mac.
• Unclamped advice exists here too: PF_REC_MEM_GB=16 (L28) → "raise to 16 GB" on Macs that don't have it (L238).
• Linux is fine (docker MemTotal == host).

Suggested widened AC: on every OS always print two labeled lines (host RAM, container-runtime budget) and clamp all recommendations at physical − 2 GB. macOS floor enforcement + VM auto-sizing is tracked separately in #428.