[compiler] Version up FlatBuffers to 23.5.26

[hseok-oh]

What

Let's upgrade compiler's FlatBuffers version from 2.0 to 23.5.26.

Issue

Compiler's quantization is used on runtime for on-device compilation and it uses different build process to make host flatc for source code generation on cross build.
So we need to disable on-device compilation build temporarily until version up is finished.

Step

• Introduce FlatBuffers 23.5.26 cmake #12905
• Disable on-device compilation build on runtime #12906
• Compiler's FlatBuffer migration
• Enable on-device compilation build on runtime #12927

To-do

Find better way to build compiler & runtime common modules

[hseok-oh]

All done. Now flatbuffers 2.0 is used for tensorflow lite old version build.