Looking at eq. (1.1) in the previous page of the book $m\cdot L$ is the number of matches, "per one unit of time" as the author writes. This is a discrete-time setup. Then $m \equiv \frac {mL}{L}$ is a straightforward proportion "number of matches divided by number of workers seeking to work, per one unit of time".
It can very well represent a probability per one unit of discrete time. Formally,
$$Prob (\text{match per unit of time}) = m $$
What the author wants is to consider continuous time. Continuous time is a difficult concept. Still, think about breaking the time-interval (it may be a year or a day) that has been "standardized" to have length $1$, in "infinite" small intervals (say milli-seconds), each having essentially zero length but all together summing up to unity (I said it is a difficult concept).
We represent these infinitesimal intervals by $\delta t$ or $dt$. So this symbol is never allowed to be "large enough", it is explicitly used to denote only something infinitesimal, of virtually no length. Then the probability in this almost non-existent interval is the corresponding fraction of the probability for a whole unit of time, so naturally,
$$Prob (\text{match per infinitesimal length of time}) = m \cdot dt$$
Summing up these minuscule probabilities over the domain, i.e. integrating in $[0,1]$ we verify this
$$\int_0^1 m dt = m\int_0^1 dt = m\cdot (1-0) = m$$