Concentration of a sparse Bayesian model with Horseshoe prior in estimating high-dimensional precision matrix