On the first one, from Cassels. He has characteristic not two. He does not say that all forms can be diagonalised over a field, he says that there is a normal basis. We use the polarization identity to get an inner product $\phi,$ a normal basis has $\phi(u_i, u_j) = 0$ when $i \neq j.$ Therefore the form applied to a vector becomes $\sum a_i x_i^2.$ Over an algebraically closed field, we can take all but the first two $x_i$ zero, then take $x_2 = 1,$ and solve for $x_1$ in $a_1 x_1^2 + a_2 = 0.$
The dimension 5 bit is a longer story. I recommend Cassels because he is focused on the rational numbers and the rational integers, he does isotropy for the $p$-adics in great and understandable detail. This includes tables for the Hilbert norm residue symbol, pages 43 and 44, with which the isotropy results are proved, pages 58-60 primarily. Your question is Lemma 2.7, page 60. The proof is pigeonhole on the $p$-adic squareclasses