Websauna should automatically retry views if a retryable exception occurs. However, this is currently bugged: Websauna won't retry views where a psycopg2.extensions.TransactionRollbackError is thrown (ie. when a serialization error occurs in PostgreSQL)

This can be tested manually by following instructions in this repo: https://github.com/koirikivi/websauna_retry_test

The first commit removes call to request.tm.abort() in websauna.system.core.views.internalservererror.internal_server_error. This call should not be needed, as pyramid_tm should handle the rollback in the case of an exception.

The second commit amends websauna.system.core.views.internalservererror.internal_server_error to avoid spamming logger.exception when a view is retried. I'm less sure about the implementation here, though it seems to work empirically. Review kindly requested!

Full story:

• Websauna uses pyramid_retry to handle view retrying
• pyramid_retry needs request.exception to retry
• request.invoke_exception_view sets request.exception but only if
  it returns a view. If no exception view is configured,
  request.exception will be None.
• Websauna adds a default exception view
  (websauna.system.core.views.internalservererror.internal_server_error),
  However, this is disabled in development by default since debug toolbar is
  enabled by default. In production settings, debug toolbar is disabled and
  pyramid_retry SHOULD work.
• However, calling request.tm.abort() kills the current zope transaction,
  which causes a new transaction to be created when the retry process happens.
• This new transaction no longer has the SessionDataManager from
  zope.sqlalchemy attached, which means that it no longer knows how to retry
  certain database exceptions, like those raised by psycopg2 when a
  serialization error occurs.