Allow precision loss on multiplying decimal arrays

[viirya]

Is your feature request related to a problem or challenge? Please describe what you are trying to do.

Current multiply kernel cannot allow precision loss when performing on decimal arrays. However, this is a requirement for decimal multiplication in practice, e.g. Spark supports it. It is difficult to ask users to manually adjust input data.

Describe the solution you'd like

It can image that such operation is slower than current multiply, so making it as default seems not acceptable. It seems more reasonable to provide additional decimal specific multiply kernel for that.

Describe alternatives you've considered

We cannot do this.

Additional context

[tustvold]

label_issue.py automatically added labels {'arrow'} from #3690