fix(stats): distinguish configured scopes from observed DB scopes

[Garyko0730]

Summary

This PR fixes a confusing scope stats mismatch in memory-pro stats / memory stats tool output.

Previously, the plugin mixed two different scope views:

• configured scopes from ScopeManager.getStats()
• observed scopes in DB from store.stats().scopeCounts

This could produce misleading output such as:

• memory.scopeCounts.agent:main = 225
• but scopes.totalScopes = 1

which looks like a broken memory system even when the database is healthy.

What changed

1. Separate configured vs observed scope stats

• added summarizeScopesByType(scopes: string[]) in src/scopes.ts
• kept ScopeManager.getStats() focused on configured scopes
• derived observed scope stats from Object.keys(stats.scopeCounts)

2. Improve CLI stats output

Updated memory-pro stats to show:

• configured scopes
• observed scopes in DB
• observed scopes by type
• configured scopes by type
• memories by scope
• memories by category

3. Improve tool/API stats output

Updated memory stats tool output/details to expose both:

• configured
• observed

while preserving top-level compatibility fields.

4. Add regression coverage

Added:

• test/scope-stats-regression.test.mjs

Updated:

• test/cli-smoke.mjs

Why this matters

Users can now distinguish:

• what scopes are declared in config
• what scopes are actually present in LanceDB

This makes debugging much easier after migrations, imports, or historical scope writes.

Validation

Ran locally:

• node --test test/scope-stats-regression.test.mjs
• node test/cli-smoke.mjs

Both passed.

Notes

This PR intentionally does not include local memory data cleanup.
Only reusable plugin code + regression tests are included.

[AliceLJY]

The configured-vs-observed scope split is the right fix here. I re-ran node --test test/scope-stats-regression.test.mjs and node test/cli-smoke.mjs on the PR branch; both passed.

[rwmjhb]

LGTM on the code. This PR currently has merge conflicts, so please rebase onto master and resolve them before merge.