[SPARK-12837][SPARK-20666][CORE][FOLLOWUP] getting name should not fail if accumulator is garbage collected

## What changes were proposed in this pull request?

After https://github.com/apache/spark/pull/17596 , we do not send internal accumulator name to executor side anymore, and always look up the accumulator name in `AccumulatorContext`.

This cause a regression if the accumulator is already garbage collected, this PR fixes this by still sending accumulator name for `SQLMetrics`.

## How was this patch tested?

N/A

Author: Wenchen Fan <wenchen@databricks.com>

Closes #17931 from cloud-fan/bug.