The concept of the bandwidth has several definitions.
For simple filters, one definition of bandwidth is the -3 dB frequency (for a lowpass filter) $f_c$ where the filter gain at $f_c$ is $H(f_c) = H(0)/\sqrt(2)$; i.e. $f_c$ is the half-power or cut-off frequency. This definition could be applied for a bandpass filters too but not for highpass filters. Note again that this definition of bandwidth requires that the filter frequency response be a continuous waveform, otherwise the the frequency will be undefined.
For signals, a 3dB bandwidth concept will not work and instead something like a 90 % energy (or power) definition is invoked. In this case the frequency spectrum of the signal is used to find the minimum frequency $f_b$ at which 90% of the total enegy of the (baseband, real) signal is contained.
In your case you have a signal with a rectangular frequency spectrum, which defies a 3 dB bandwidth, but enables the computation of a 90 % bandwidth definition.
However note that all these definitions make sense under typical waveform assumptions, but for your case neither -3dB nor 90% enegy definitions adequately define the signal's bandwidth.
You signal is bandlimited in the sense that the ractangular function is exactly zero out of its lobe and that frequency shall be taken as the bandwidth of this signal.