You'll have to write the code in C to change the GPIO configuration to Open Drain w/Pull-Up if you don't have external ones.
You'll then need to sequence writes to the SCL/SDA pins to emulate the data/clocking expectations in the data sheet.
Might this conflict with the other I2C devices? Perhaps, I don't know these are my parts. You'll have to experiment.
Would it be simpler to switch back and forth between I2C and bitbanged mode? Again this is a call you'll have to make, clearly bitbang mode would be more involved, but you could code it to recognize which device you're addressing, and change the address and data widths to match expectations of each device on the bus. Then you'll need to determine if they play well or not. Data sheet's probably not going to tell you either way. You'd see if they NACK as usual to their own addresses, and that affirms they are decoding their own address properly.
Will software drop out of a Google search? There's probably some examples of bit bang I2C, the VCSEL perhaps? The concept of driving the pins in a more direct fashion is not a new one. Should be some close cousins you can adapt.
