Latency of 10GBASE-SR/-LR vs SFP+ DAC is very closely the same - in contrast to 10GBASE-T which adds appr. 1.5 μs.
Majorly, latency is caused by the line encoding overhead. In the case of -R PHYs, that's 64b66b code which requires little processing. SFP+ cages are directly fed with an -R data stream, which DACs then couple onto copper twinax and SFP+ modules use to modulate their lasers. Again, very little processing, same low latency for both.
Hypothetically, DAC _could_ be faster because of the potentially higher velocity factor (VF) and, accordingly, lower propagation delay. The kind of twinax used is at the vendor's discretion, but even if that had a VF of .9 in comparison to fiber's .67, the delay difference for a long DAC of 10 m against fiber would be around 12 ns (10 / c * [1/VFfiber \- 1/VFtwinax]).
It makes more sense to tailor the patch cables than to worry about that (12 ns is about 2.5 m worth of fiber) but either is rather moot.