Kai: result tables float within the evaluation section, which flushes them before the limits.

Co-authored-by: Hanzo Dev <[email protected]>

kai/sections/08-evaluation.tex

@@ -43,7 +43,7 @@ and is read as such.
 \subsection{Accuracy on the frozen harness}
 \label{sec:accuracy}
 
-\begin{table}[t]
+\begin{table}[tbp]
 \centering
 \small
 \setlength{\tabcolsep}{4pt}
@@ -114,7 +114,7 @@ kept runs are at \texttt{4e57fb2}.
     to a third party \meas.
 \end{description}
 
-\begin{table}[t]
+\begin{table}[tbp]
 \centering
 \small
 \setlength{\tabcolsep}{4pt}
@@ -141,7 +141,7 @@ reference, Laya's kept Python run on an M1 Max took 53\,ms at $N = 1$ and
 43.9\,s at $N = 1{,}000$, and Jev 203\,ms and 730\,ms at \$0.0126 per
 1{,}000-question call \meas{} (\texttt{4e57fb2}).
 
-\begin{table}[t]
+\begin{table}[tbp]
 \centering
 \small
 \setlength{\tabcolsep}{4pt}
@@ -151,3 +151,6 @@ questions per state, on Metal and on CPU. Written by \texttt{bench tex}; the
 last row names the machine and revisions.}
 \label{tab:speed}
 \end{table}
+
+% Flush the result tables here, so none drifts past the section.
+\clearpage