This is an interesting question. I've recently studied the lightning protection practices for radio towers for another project, so the first thing that comes to my mind is whether you need to consider lightning protection or not. The frame is external to the building, but directly against it, so I would think that the building (and perhaps neighboring structures) provides an envelope of protection against direct lightning strike. However, the chances of a lightning surge current coupling to the frame from the building is dependent on how well the building's internal steel is grounded and what kind of lightning protection the building has. On the other hand, San Francisco is not a high lightning probability area anyways. Is this a permanent installation?
The threat of RF voltages and currents depends on a few things. First, I'm wondering what electrical items may react to RF as this tells us what we need to protect. I don't think the function of the LEDs themselves can be affected by RF voltage or current. I do seriously believe that the device that drives the LED feed wires, that is some sort of controller circuit, will definitely need protection from RF at the point where it interconnects to the array. This would follow standard practice in dealing with common mode and differential RF voltages and currents, which is beyond the scope of this reply.
There is also two mechanisms by which you are also likely to cause radio interference. One is the small possibility that your LEDs will rectifiy RF that is induced into the wiring system, and then re-radiate the result. This is a classic intermod generator. The threat mainly exists for very brief times when you are switching an LED on or off, but with so many LEDs I think that, statistically, this may become significant. The other mechanism that we have to consider by which you might cause radio interference is that your wiring will radiate switching transients out into the radio spectrum and pollute the airwaves. I think that you should design to meet FCC part 15 rules for radiation, as a bare minimum.
As for whether you should attach the frame to the building and how, I think the first criteria is safety. Single point grounding is a common concept when wanting to minimize the effects of surge currents. Multiple point bonding would create circulating currents and this seems like a bad idea. I'm a bit out of my depth on this, so perhaps we need to hear from an engineer with more familiarity in the principles of grounding large structures. On the other hand, when you have one conductor running beside another you can have differential surge voltages that cause arcing between them, so for that reason, my gut says to bond your frame to the building in several widely spaced points. While this adds metal volume behind your frame, I don't think that this makes your frame a larger antenna, at least not by much. The effective aperture of the frame alone, vs the frame intimately bonded to the building doesn't seem all that different to me.
Regardless of how the safety bonding is done, you should prepare your driver circuits to repel RF using appropriate filters on all wires that lead up to the LEDs. So, now I'm wondering what the wiring topology will be. Will be creating a large matrix? How will you drive the LEDs?