[non-BE] Resolve "Low bitrate SBA tuning fix"

Related to #487 (closed)

The fix here is to reduce the worst case SCE bitrates by 150bps to accommodate the theoretical worst case SPAR MD bitrate (including 11 bits of signalling overhead). This fix should be BE for most inputs, barring individual frames where the SPAR MD bitrate was very close to the limit. We tested our 16 regular low bitrate FOA inputs, and found 13/16 were BE. The three vectors that were not BE were different for a handful of frames around a frame where a slightly different decision was made (in one case for Priority::Trivial frames, and in the other two much less), with bit differences quickly dropping to <= 3 LSB for the remainder of the vector. PEAQ showed almost identical results for the non-BE vectors.

On CI, the only non-BE tests are for SBA bitrate switching cases that include 13.2 and 16.4kbps. The differences are clustered around (some of) the frames transmitted at these bitrates.