enforce database timeouts

[HayimShaul]

Set a timeout to database queries to protect against database queries that are maliciously designed to take a long time.
Addresses issue #1633

[AkramBitar]

The query timeout issue is tracked also under issue: #1740
This issue suggest to add the query timeout in the database level as a configuration parameter, while other solution can be in the application level like this PR suggests.

The timeout of the query can be defined for PostgreSQL in the database level but for SQLite we do not have this option.

Need to discuss with @adecaro and @mbrandenburger which approach is better.

PostgreSQL:

ALTER ROLE app_user
SET statement_timeout = '30s';

Go:

ctx, cancel := context.WithTimeout(ctx, 30*time.Second)
defer cancel() 

[AkramBitar]

The query timeout issue is tracked also under issue: #1740 This issue suggest to add the query timeout in the database level as a configuration parameter, while other solution can be in the application level like this PR suggests.

The timeout of the query can be defined for PostgreSQL in the database level but for SQLite we do not have this option.

Need to discuss with @adecaro and @mbrandenburger which approach is better.

PostgreSQL:

ALTER ROLE app_user
SET statement_timeout = '30s';

Go:

ctx, cancel := context.WithTimeout(ctx, 30*time.Second)
defer cancel() 

After consulting with @adecaro,
We need to make sure that token sdk respects the context timeout.
In case we have a context timeout (or when the context canceled) then the query should be interrupted ASAP.

[AkramBitar]

@HayimShaul
The context is indeed passed through to ExecContext(ctx, ...)/QueryContext(ctx, ...), so cancellation can propagate to the driver.

Should we add a test that runs a slow blocking query (e.g. pg_sleep/SLEEP or a fake blocking driver) and asserts that QueryContext/ExecContext is actually interrupted and returns context.Canceled or context.DeadlineExceeded instead of hanging?