feat(database): disable eager collection creation for all schemas

[ChrisPdgn]

Summary

• Set autoCreate: false and autoIndex: false on every Mongoose schema in the database adapter, not just views.
• Collections are now created lazily on first write instead of at schema registration time, reducing empty collections on shared dev MongoDB clusters.
• Index creation remains handled explicitly by Conduit's createIndexes() during schema registration.

Test plan

• Start Conduit with MongoDB and register schemas that have no custom indexes; verify no empty collections are created until the first document insert.
• Register a schema with custom indexes and confirm indexes are still created correctly.
• Run findMany / findOne against a schema whose collection has not been created yet and confirm empty results are returned without errors.
• Verify view schemas still behave correctly.

[ioanniskemerlis]

Index creation remains handled explicitly by Conduit's createIndexes() during schema registration.

That’s only true for indexes that are still left in modelOptions.indexes after schemaConverter runs. Single-field indexes get peeled out of modelOptions.indexes and moved onto the field (SchemaConverter.convertModelOptionsIndexes), and field-level unique / index never go through createIndexes() at all. They used to rely on Mongoose autoIndex.

So with autoIndex: false on every schema, those indexes aren’t created on a fresh MongoDB. Compound ones still work.

Quick examples:

ChatRoom: { fields: ['participants'] } gets moved to the field; only { fields: ['participants', 'deleted'] } hits createIndexes()
EmailTemplate.name with unique: true, no unique index gets created on register

I checked this with Mongoose too: with autoIndex: false, path indexes show up in schema.indexes() but tagged _autoIndex: false, so they won’t be built. There’s also no ensureIndexes / syncIndexes fallback elsewhere in the adapter.

This won’t show up on DBs that already had the indexes from before, but it’s a real regression for fresh environments / new schemas.

I think we should keep autoCreate: false, but either:

• explicitly create field-level indexes after model init, or
• stop moving single-field indexes onto fields and always create them via createIndexes() (closer to the SQL path)

[ChrisPdgn]

Summary of changes

This PR disables eager MongoDB collection creation in the Mongoose adapter while preserving correct index behavior on fresh databases.

1. Lazy collection creation

In MongooseSchema.ts, all schemas now use:

• autoCreate: false
• autoIndex: false

This stops Mongoose from creating empty collections at schema registration time. Collections are created lazily on first write, or when an index is explicitly created.

2. Explicit field-level index creation

Disabling autoIndex exposed a regression: field-level indexes (unique: true, index: {...}) and single-field indexes moved onto fields by SchemaConverter were no longer being created on fresh MongoDB instances.

To fix that, MongooseAdapter now explicitly creates indexes declared on the compiled Mongoose schema via a new createMongooseFieldIndexes() helper, called from _createSchemaFromAdapter() after the existing createIndexes() path.

Important safeguards:

• Skips the default { _id: 1 } index so schemas with no other indexes remain truly lazy
• Skips views
• Strips Mongoose-internal _autoIndex metadata before calling MongoDB createIndex()
• Preserves existing compound-index creation through createIndexes()

Expected behavior

• Schemas with no custom indexes: no collection created until first insert
• Schemas with field-level indexes: collection + indexes created at registration
• Schemas with compound indexes: unchanged, still created via createIndexes()
• Reads against non-existent collections: return empty results without error

No Conduit schema migration is expected; _DeclaredSchema stores the original schema, not the converted Mongoose runtime schema.