MEDIAN() aggregation support

N1QL is added following aggregates in Mad-hatter ( MB-16168)

*    MEDIAN( x),             MEDIAN(DISTINCT x)

*    STDDEV( x),             STDDEV(DISTINCT x)

*    STDDEV_SAMP(  x),         STDDEV_SAMP(DISTINCT x)

*    STDDEV_POP( x),           STDDEV_POP(DISTINCT x)

*    VARIANCE( x),          VARIANCE(DISTINCT x)

*    VAR_SAMP( x),         VAR_SAMP(DISTINCT x)

*    VAR_POP( x),            VAR_POP(DISTINCT x)

Out of this only Some cases of MEDIAN() can be pushed to indexer, all others can't be pushed due to computation required to hold all the data.

In case MEDIAN(), Push down can only happens same rules that we apply for DISTINCT aggregates

• Full aggregate
• Aggregate argument must be exactly matched with one of n+1 group keys
• Partition keys must be in group keys

If supporting the MEDIAN() also not possible due to holding data (further think it requires hold data), mention as resolution, we will not pushdown