Add one blog post

[felix-yuxiang]

Update a blog post on diffusion distillation.

[felix-yuxiang]

Some more issues to be fixed. Suggested by @qiyan98

• Zero-tangent loss: position of trainable model, when should it be inside the total derivative or outside? Namely, how should we apply JVP to the trainable/offline model?
• explain the details of l_2 with cosine loss?
• How does AYF convert to vanilla FM loss when s->t? mention s->t is indicating taking a limit
• MMD loss inside IMM, are both outputs of $f$ functions differentiable?