Add new check for indexing with atomic=True

Briefly, this check helps prevent a long-running index creation from
preventing all reads and writes to a table.

This supplements an existing check
==================================
There is an existing check, `CREATE_INDEX` which warns about creating
indices without the `CONCURRENTLY` argument. It's a helpful check!

However, there are some valid reasons to prefer non-concurrent index
creation. For example, concurrent index creation may struggle to
complete when applied to write-heavy tables (since Postgres will have to
make repeated passes on the table to accommodate the new reads since the
last round).

The danger of `EXCLUSIVE` locks
===============================
If you *do* want to create indices nonconcurrently, there is still a
**huge footgun** to be aware of -- locks which are held in a transaction
are not released until the transaction closes.

Dangerous Django defaults
=========================
By default, any time you add a new Django field with `db_index=True` or
`unique=True`, the automatically generated `Migration` has:

- implicit setting of `atomic=True` (transactions are on)
- an `ALTER TABLE...ADD COLUMN` call (exclusively locks the table)
- a `CREATE INDEX` invocation (*hopefully* quick, due to nullable
  columns)

It's even more dangerous if you combine an `ALTER TABLE` call with
`CREATE INDEX` of an existing, non-nullable column.

Favor `atomic=False` if you must index
======================================
Index-creating migrations should have few `operations`. Since index
creation does not modify row-level data (and only requires locking
against writes), there's no reason to use transactions.

It's prudent to bypass the default `atomic=True` for index creation.

django_migration_linter/sql_analyser/postgresql.py

@@ -5,6 +5,36 @@ import re
 from .base import BaseAnalyser, Check, CheckMode, CheckType
 
 
+def has_create_index_in_transaction(sql_statements: list[str], **kwargs) -> bool:
+    """Return if a migration opens a transaction, acquires EXCLUSIVE lock, then indexes.
+
+    Any locks that are obtained after a transaction is opened will not be released
+    until that transaction either commits or rolls back.
+
+    Accordingly, it's extremely important that a migration avoid any long-running
+    queries while exclusive locks are held.
+
+    Building an index is generally a long-running operation. A standard index
+    creation (that is, avoiding use of `CONCURRENTLY`) can be safe to run on
+    tables which are small in size and/or read-heavy. (`CREATE INDEX` only
+    prevents writes to a table; reads are allowed during index creation).
+
+    This check is a stricter version of `CREATE_INDEX` -- if a team wishes
+    to build indices nonconcurrently, it's imperative to be mindful of locks.
+    """
+    if not (sql_statements and sql_statements[0].startswith("BEGIN")):
+        return False
+
+    for i, sql in enumerate(sql_statements):
+        # If any statements acquire an exclusive lock, complain about index creation later.
+        # Nearly every single `ALTER TABLE` command requires an exclusive lock:
+        #     https://www.postgresql.org/docs/current/sql-altertable.html
+        # (Most common example is `ALTER TABLE... ADD COLUMN`, then later `CREATE INDEX`)
+        if sql.startswith("ALTER TABLE"):
+            return has_create_index(sql_statements[i + 1 :])
+    return False
+
+
 def has_create_index(sql_statements: list[str], **kwargs) -> bool:
     regex_result = None
     for sql in sql_statements:
@@ -32,6 +62,13 @@ class PostgresqlAnalyser(BaseAnalyser):
             mode=CheckMode.TRANSACTION,
             type=CheckType.WARNING,
         ),
+        Check(
+            code="CREATE_INDEX_EXCLUSIVE",
+            fn=has_create_index_in_transaction,
+            message="CREATE INDEX prolongs transaction, delaying lock release",
+            mode=CheckMode.TRANSACTION,
+            type=CheckType.ERROR,
+        ),
         Check(
             code="DROP_INDEX",
             fn=lambda sql, **kw: re.search("DROP INDEX", sql)