Change SQL fomatting tool to pgformatter (#477)

Change the SQL formatting tool used by `make format` from
[sql-formatter](https://github.com/sql-formatter-org/sql-formatter) to
[pgFormatter](https://github.com/darold/pgFormatter).

Stop using `sql-formatter` because it doesn't work on the `read_schema`
function:
 
```
npx sql-formatter -l postgresql --fix pkg/state/init.sql
````  
doesn't change the file regardless of whatever changes are made to
`read_schema`.

`pgFormatter` produces consistent formatting on the `read_schema`
function so let's use that.

I'm not convinced that either formatting tool gave us formatting that
was better than what we were doing by hand before. It's a question of
consistency vs personal preference.

Author: andrew-farries

.github/workflows/build.yml

@@ -73,6 +73,7 @@ jobs:
 
       - name: Generate types
         run: |
+          chmod 666 pkg/state/init.sql
           make generate
 
       - name: Ensure generated types are up-to-date

Makefile

@@ -10,7 +10,7 @@ format:
 	# Format JSON schema
 	docker run --rm -v $$PWD/schema.json:/mnt/schema.json node:alpine npx prettier /mnt/schema.json --parser json --tab-width 2 --single-quote --trailing-comma all --no-semi --arrow-parens always --print-width 120 --write
 	# Format embedded SQL
-	docker run --rm -v $$PWD/pkg/state/init.sql:/mnt/init.sql node:alpine npx sql-formatter -l postgresql -o /mnt/init.sql /mnt/init.sql
+	docker run --rm -v $$PWD/pkg/state/init.sql:/data/init.sql backplane/pgformatter --inplace /data/init.sql
 
 generate: format
 	# Generate the types from the JSON schema