SARVLM: A Vision Language Foundation Model for Semantic Understanding and Target Recognition in SAR Imagery