Board Support for the stm32mp1 platform is contained in the device tree,
so if we remove hardcoding of board name from the Makefile, we can build
the intermediary objects once and generate one new tf-a-*.stm32 binary
for every device tree specified. All in one go.

With implicit rules implemented, we only need to change the top level
target to support multi-image builds on the stm32mp1.

This makes BSP build system use of stm32mp1 TF-A more convenient, because you need not duplicate recipe and build to support, say, the discovery kit and an own hardware based on it.

(If I understood correctly, I am supposed to have an issue for patches I send. I will send the patches shortly)