Abbreviation period displaced onto the following number instead of the abbreviation (Nov .8 for Nov. 8)

When an abbreviation (\`Nov\`, \`No\`, \`Figs\`, \`pp\`, ...) is immediately followed by a number, the period ends up attached to the *number* instead of the abbreviation: \`Nov .8\` for Brown's \`Nov. 8\`, \`No .3\` for \`No. 3\`.

Detection: a bare-letters token immediately followed by a single space then a \`.\` + digits token. Most matches for this shape are ordinary decimal numbers with a real word before them (\`batted .365\`, \`the .028\`) and are already correct -- confirmed individually against \`src/semcor/brown-nolines.txt\` (both shapes are genuinely present in Brown's real text). Only the 12 instances where the preceding word is itself a recognized abbreviation (\`Nov\`/\`Oct\`/\`Sept\`, \`No\`, \`Figs\`, \`pp\`) are real bugs, each confirmed individually.

Fix is a 2-character swap (the gap and the following period trade places), moving the period onto the abbreviation and leaving a single space before the number -- token count unchanged, only the two adjacent tokens' spans shift by one character each.

Inspect the 12 confirmed cases in src/semcor/brown-nolines.txt, focusing on the recognized abbreviations Nov, Oct, Sept, No, Figs, and pp. Correct the adjacent gap and period while preserving the token count, then verify that each affected abbreviation and following number has the intended spacing and spans.

Written by the indexing model from the issue text.