Created by: wojtuss

With this patch in QAT:

• the --quantized_ops option was renamed into --ops_to_quantize and made optional; if the option is not used, an attempt to quantize all quantizable oprators is made,
• if quantization scale is missing for any quantized operator, the quantization of the op is skipped with a message in the log, without error,
• QAT INT8 test for Ernie does not use the --ops_to_quantize option, i.e. all quantizable ops are quantized,
• the algorithm is refactored,
• QAT doc is updated.

test=develop