How to make a continuous fill legend appear discrete

Question

I am using gganimate with colorspace to demonstrate partial pooling in Bayesian hierarchical models, and I like the aesthetic transition I have accomplished, but I would prefer for the fill legend for "Prior" to appear as two discrete boxes (as it would with scale_fill_discrete) but I seem to need to pretend my distribution field is continuous in order to trigger the nice transition that scale_fill_continuous_sequential() is enabling.

library(tidyverse)
library(colorspace)
library(gganimate)
library(ggplot2)

set.seed(42)
SIMS <- 1e3
df <- 
    tibble(
        a = rbeta(SIMS, 600, 400),
        b = rbeta(SIMS, 650, 350)
    ) |> 
    pivot_longer(
        everything(),
        names_to = 'distribution',
        values_to = 'values'
    ) |> 
    mutate(distribution = factor(distribution, levels = c('a','b')))

df |> 
    ggplot(aes(
        x = values, 
        fill = as.integer(distribution)
        )) + 
    geom_density() + 
    colorspace::scale_fill_continuous_sequential(
        palette = "Blues 3", 
        begin = 0.1, end = .9,
        breaks = c(1,2)
        ) +
    transition_states(
        states = distribution,
        transition_length = 1,
        state_length = 1,
        wrap = TRUE) + 
    labs(fill = "Prior")