I don't know of any way to bypass the tape loading. I do all my development on a Linux box I VNC into, editing and assembling there using command line tools. I then use the web intf to load the reader, the virtual panel on the Linux box to read it in. That seems to be the simplest way, at least for me.
Theoretically, I suppose someone could write a standalone program to directly copy the contents of a tape, decoded from rim/binloader into the actual image, into the running Pidp-1. I believe its context, including memory, is accessible via shared memory, but I'm not sure. Probably more work than justified.
Bill