Move registrySupportsTimeField into registryOptions as a per-registry setting

[zkochan]

Summary

registrySupportsTimeField is a single global boolean, but it describes a capability of an individual registry: whether its packument carries the time field. In a workspace that resolves from more than one registry — the public npm registry plus a self-hosted or mirrored one — one flag cannot be right for both.

Now that #13942 adds registryOptions, a per-registry map keyed by registry URL, this belongs there as supportsTimeField.

Why it matters today

Two code paths read the flag once and apply the answer to every registry:

• pnpm11/cache/commands/src/cache.cmd.ts:69 — picks the cache type when resolutionMode is time-based.
• pnpm11/exec/commands/src/dlx.ts:115 — the same decision for dlx.

A workspace whose default registry serves time but whose scoped registry does not (or the reverse) gets the wrong behavior for one of them, with no way to express the difference. minimumReleaseAgeIgnoreMissingTime is the same registry deficiency seen from the other side — what to do when time is absent — so the two should be looked at together; it may be that one per-registry fact subsumes both rather than moving each as-is.

Proposed shape

registryOptions:
  https://npm.internal.example.com/:
    supportsTimeField: false

The existing global registrySupportsTimeField stays as the default for registries with no entry, so nothing changes for anyone who does not opt in. The registry prefix in the name becomes redundant once it sits under a registry key, which is a small bonus.

Notes

• This is a correctness fix for multi-registry workspaces rather than a refactor; framing it that way should decide whether it is worth the churn.
• Applies to both the TypeScript CLI and pacquet, like the setting it joins.
• Depends on feat(config): declare each registry once in the registries setting #13942 landing first.
• The admission rule worth keeping for registryOptions: it holds facts about the server, not policies about your project. supportsTimeField is a fact. minimumReleaseAge and trustPolicy are policies and should stay global — per-registry granularity there would make it easy to quietly weaken supply-chain checks for one host.

---

Written by an agent (Claude Code, claude-opus-5).