firstly this depends on whether you are using 7-mode of clustered onTap with your deployment - if you are using clustered, it is possible to do a true active passive, which is all drives on the active controller.
However if we assume 7 mode, which is the most likely version based on what you've said, i'd agree with what you have said, if you are trying to maximise performance and capacity from the disks then all the disks on one controller (with the exception of the three as you have suggested to allow the "passive" node to boot) is certainly an acceptable setup when the controllers have limited spindels available.
obvioulsy this can depend on work load that you are expecting to push through the single controllers, but we'll assume that that is OK.
as for your other questions
Am I right in thinking that for these smaller filers, it's acceptable to stick the root vol in the same aggr as the data volumes? - yes it is...technical in 7 mode nothing wrong with doing that - the reasons you don't are really around recovery, if you where to loose the aggreagte containing the root vol, the smaller the aggr the quicker recovery, outside of that no reason not to.
So we could therefore use all of those 17 disks in a RAID-DP aggregate with a single/two? spare(s)? - so 2 spares is the "good practice" however in a smaller setup and if you want to make the most of the drives you have, one spare is fine - and if we are maximising capacity and performance then 17 drives in the single aggreagte makes perfect sense.
Using the NetApp space calc, that suggests the useable space with a RG of 15 will be just over 6TB with 0% Vol Snap res... OK so think carefully about RAID Groups a RAID Group of 15 maybe not ideal here - probably worth thinking logically...once you fill a RAID Group when a new on is created is needs a further two parity drives, so for example if we have 16 usable drives here - then would make sense the RAID Group is big enough to take all the drives you have available in a single RAID Group - but also think about expansion in the future, so you don't get unbalanced groups- for now i'd probably make it a group of at least 16 then maybe if you add a shelf in future go for a new aggregate and size accordingly - for now this will give you 14 data drives of 540gb (or whatever the right sized value of the 600gb drive is)
Or if we need a seperate 2 disk root vol aggregate, 15 disks in RAID-DP with a RG of 13 will be just over 5.1TB usable... probably wouldn't go down this route... you can't really do a two disk root aggregate, unless you are going RAID 4 with it... which i'd suggest you don't you;d then only be left with 13 drives for the aggregate and only 11 for data...so losing a further 3 drives worth of capacity.
hope all that helps... you are of course completely right to think about it... and ask the question now...
sounds like you may be UK based, so if you are and struggling, let me know, happy to help if I can.
the comments about goign CDOT is a very valid one, and should be considered... using 8.3 on the 2552 and root disk partitioning, potentially gives you a much better solution, with all of the NetApp innovation going into CDOT only, 7-mode operations will remain pretty much as is... it will be supported for a long time to come, but as a platform the CDOT stuff gives you a whole lot of benfit and flexbility into your environment.