Check if it's worth allowing changing the builders for strict Text/strict ByteString
We currently use lazy Builders as the backing for both Text and ByteString interpolation, in order to make sure that we can scale to very large interpolations. However, for applications that primarily interpolate lots of small strings, the overhead implied by that might be an unnecessary bottleneck.
We should run some benchmarks to see if changing the backing, perhaps gated by a Cabal flag, improves performance for these scenarios.