files.id to 64 bits

Seeding adds ~14M files a day; 32-bit ids run out at 2.1 billion in about
five months. Postgres can't change a partition key's type, so the
migration detaches the partitions, rewrites each as a plain table (a CHECK
added first spares the attach scan), and attaches them to a new 64-bit
parent with the same indexes, foreign keys and triggers. binaries.file_id
and the id sequence become 64-bit too. Meant to run with the writers
stopped and Postgres on wal_level=minimal (compose: PG_WAL_LEVEL,
PG_MAX_WAL_SENDERS).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

database/schema.ts

@@ -40,7 +40,7 @@ export class BinarySchema extends BaseModel {
   @column.dateTime()
   declare conflictedAt: DateTime | null
   @column()
-  declare fileId: number | null
+  declare fileId: bigint | number | null
   @column()
   declare fileKey: string
   @column.dateTime()
@@ -192,7 +192,7 @@ export class FileSchema extends BaseModel {
   @column()
   declare groups: any | null
   @column({ isPrimary: true })
-  declare id: number
+  declare id: bigint | number
   @column()
   declare parts: number
   @column()
@@ -229,7 +229,7 @@ export class FilesP0Schema extends BaseModel {
   @column()
   declare groups: any | null
   @column({ isPrimary: true })
-  declare id: number
+  declare id: bigint | number
   @column()
   declare parts: number
   @column()
@@ -266,7 +266,7 @@ export class FilesP1000000Schema extends BaseModel {
   @column()
   declare groups: any | null
   @column({ isPrimary: true })
-  declare id: number
+  declare id: bigint | number
   @column()
   declare parts: number
   @column()
@@ -303,7 +303,7 @@ export class FilesP5000000Schema extends BaseModel {
   @column()
   declare groups: any | null
   @column({ isPrimary: true })
-  declare id: number
+  declare id: bigint | number
   @column()
   declare parts: number
   @column()