[BP 7.2.7] XDCR - Recompression is missed in mobile mode before replicating the mutation

In the mobile mode, XDCR will need to decompress data to read xattrs section and update HLV etc, if the body received is in a compressed format.

Right now, to re-compress the decompressed body before replicating, we have the following logic in mobile mode:

https://github.com/couchbase/goxdcr/blob/master/parts/xmem_nozzle.go#L1476

The body is recompressed only if:

maxEncodedLen := snappy.MaxEncodedLen(len(mc_req.Body))

if maxEncodedLen > 0 && maxEncodedLen < len(mc_req.Body) {

      // compress mc_req.body

}

But a positive maxEncodedLen is always: (Refer https://github.com/golang/snappy/blob/v0.0.4/encode.go#L78)

32 + len(mc_req.Body) + len(mc_req.Body)/6 

which is always greater than len(mc_req.Body)

So the decompressed body is never re-compressed.

A potential solution could be:

1. Keep note of the pre-decompressed body length

2. Decompress body, keep note of the new body length

3. Update xattrs (HLV etc), use the same doc body -> updateSystemXattrsForTarget(...)

3. If length in (1) < length in (2), compress it back [Since we don't modify the doc body and assume xattrs are compressible] 

A side note: Snappy compression is only supported for *_with_meta, and this should be skipped for subdoc command.

The change should also be such that it honours the XDCR replication setting for compression - https://docs.couchbase.com/server/current/xdcr-reference/xdcr-advanced-settings.html#xdcr-compression-type-values