glowell: did you decide on a solution for for your gimbal system? This is probably the wrong forum to post these questions on. Better to check out rcgroups, diydrones, openpilot forums and even 9xforums.
I'm an APM2.5 user, but I have yet to see much success from people using it's Camera Gimbal system for trying to shoot stabilized video. One thing to note is the APM2.5's two PWM outputs meant for gimbal servos (pitch & roll) only output at 50Hz. This is far too slow for high speed digital servo's, which operate at up to 333Hz. According to Askman, who's a fairly renowned gimbal designer, you ideally want your update rate to be double your servos. The HoeryFly Gimbal board supports an update rate of 800Hz (i think). That said, I believe it's possible to use two motors outputs instead for Gimbal Control on the APM. the motor outputs output at 400Hz. I haven't tried this myself.
I'm currently in the middle of building this Gimbal kit from Rusty'shttp://www.shop.aglhobbiesllc.com/Camera-Mounts/UG-1-Full-Kit-2-axis-gimbal.html
It's the result of lots of work by Rusty and Hansen, who was a long time beta tester of Askman's gimbals. You can follow it's development through this extremely interesting thread:http://www.rcgroups.com/forums/showthread.php?t=1739869
I've ordered two Hitec HS-7955 servos from Servocity. One modified for continuous rotation for tilt and another which will be programmed to 180° for roll. There are lots of discussion in the later pages of the thread above for cheaper alternatives, notably some very well performing MI and BL servos from HobbyKing.
Regarding a standalone Arduino Gimbal controllers, I noticed this code a while back for employing the ArduIMU for stabilization:http://diydrones.com/group/arduimu-gimbal-controller?xg_source=activity
Though I'm not sure if anyone at all has ever used it.
I would be more inclined to simply use MultiWii if I was going to go this route..http://www.multiwiicopter.com/products/paris-gimbal-3-axis-control-board
Though apparently you only get about 70% the performance of the Hoverfly board with it. Again, I'm not yet sure how APM's gimbal system compares to either.
The issue I believe with Arduino based gimbal control is simply that the processor can't update fast enough. Controllers like Hoverfly, Wookong, AutoQuad 6.6, OpenPilot, Skyline RSGS and likely Radian all use much faster 32bit ARM processors. Arducopter is on it's way and has been successfully tested on their PX4 board, but they aren't there yet.
A lot of peeps seem to be having quite a lot of success with OpenPilot as both an FC and gimbal control system. The issue with OpenPilot though has simply been availability. I got in on the pre-order for their latest Revolution board and am excited to check out the platform. Their GSG system looks much more polished than APM's Windows only Mission Planner.
Last, is your friend building an Octo an experienced pilot? If not, let him smash up the coper for a while before you ever even consider attaching a gimbal to it. When you do attach the gimbal, get some experience flying with dead weight first before you attach a camera. I also highly highly recommend getting some flight time with some mini/micro quads. The WLtoys v929/939/949 series will all natively bind with a Turnigy 9x's default rf module. They're a blast to fly and this combo is hell of a lot cheaper than going the DSM Spektrum radio + Blade MQx route. Especially after you've flashed your 9x with er9x, you won't look back.
Oh, other fun thing to note, there are still AVR/Arduino chips in lots of copter components. Tons of ESC's for example use AVR chips, hence you can flash them with SimonK's firmware for better multicopter performance:http://wiki.openpilot.org/display/Doc/RapidESC+Database
Also MinimOSD is simply an ATmega328P paired with a MAX7456 monochrome display chip:http://code.google.com/p/arducam-osd/
And even with the mass migration to ARM based chips, Arduino isn't going anywhere thanks to the DUE. The Ardupilot team is even currently assessing using the Arduino IDE for their migration to 32bit for maintaining compatibility and the current code base as much as possible.