* [LPT] reused reference FQ implementation in fold_fake_quantize * [LPT] Removed legacy parameters * Added plugin tests with per-channel FQ for GrConv wo reshape * Apply folding only in the case when FQ data input is constant * EliminateFQ fix
* [LPT] reused reference FQ implementation in fold_fake_quantize * [LPT] Removed legacy parameters * Added plugin tests with per-channel FQ for GrConv wo reshape * Apply folding only in the case when FQ data input is constant * EliminateFQ fix