I won't go further than a very simplified (1st order) model of what's going on to help develop a meaning for \$f_t\$. There are higher order models. But to a first order the following approximation should suffice:
simulate this circuit– Schematic created using CircuitLab
Bipolars are voltage controlled devices. Not current controlled. Given the above, \$i_b=C_\beta\, s\, V_f\$ and \$i_c=g_m\,V_f\$. For the purposes of understanding \$f_t\$, it occurs when \$\vert\frac{i_c}{i_b}\vert=1\$.
Derived directly from the above, \$C_\beta=\frac1{2\pi}\frac{g_m}{f_t}\$ then defines a relatively flat relationship between \$f\$ and \$g_m\$. As \$I_\text{C}\$ (or \$g_m\$ which is proportional) increases so does \$f_t\$ in linear fashion and thus the constant \$C_\beta\$ captures this.
\$C_\beta\$ will remain relatively constant for frequencies lower than the given \$f_t\$. That's why it can be listed on the datasheet as a parameter, in fact.
Once the peak \$f_t\$ is reached, however, \$C_\beta\$ increases proportional to the square of \$I_\text{C}\$.
In a sense, \$C_\beta\$ is kind of like \$h_\text{FE}\$ which appears to be relatively constant for any given operating point. (And both aren't relatively constant when certain things like \$I_\text{C}\$ exceed a certain level.)
In your device's case, they specify that a typical value of \$f_t\$ is \$100\:\text{MHz}\$ when operating at \$I_\text{C}=100\:\text{mA}\$ and \$V_\text{CE}=10\:\text{V}\$. At room temperature, you can work out that \$g_m=\frac{100\:\text{mA}}{V_T\approx 25.9\:\text{mV}}\approx 3.85\:\mho\$. So \$C_\beta\approx 6\:\text{nF}\$ for this device.
But keep very much in mind that both \$f_t\$ and \$g_m\$ are functions of a lot of other device parameters and the operating point. So it's important to understand those additional device parameters when evaluating it. The datasheet likely selected the more optimistic operating point for the device. So it's really not possible to use this simplified result as some kind of bright line understanding of the device. It's just one data point to give you a rough idea of the best case expectation, I think.
Note also that the datasheet specified \$I_\text{C}=100\:\text{mA}\$. There's a reason. For RF work, the bipolar is often operated at higher collector currents. The reason is to achieve higher \$g_m\$ such that \$g_m\gg \frac1{R_\text{E}}\$ since, in that case, \$C_\beta\$ can be neglected in the design.
Is it the maximum switching frequency of the transistor or is itindicating something else?
To expand the above would require acquiring a lot more data on some device, then writing up examples -- first, of emitter followers; then second of other stages -- and then analyzing them using the parameter. I'm not prepared to write that.
So I stop here.