Move + to EOL instead of BOL

The leading symbol triggers tensorflow.org's markdown processor to treat it as a list, so the added linebreaks prevent mathjax from being parsed correctly.

PiperOrigin-RevId: 558802456

Author: markmcd

tensorflow_probability/examples/jupyter_notebooks/Fitting_DPMM_Using_pSGLD.ipynb

@@ -411,8 +411,8 @@
         "To update parameters $\\boldsymbol{\\theta}\\equiv\\{\\boldsymbol{\\pi},\\,\\alpha,\\, \\boldsymbol{\\mu_j},\\,\\boldsymbol{\\sigma_j}\\}$ in $t\\,$th iteration with mini-batch size $M$, the update is sampled as:\n",
         "\n",
         "$$\\begin{align*}\n",
-        "\\Delta \\boldsymbol { \\theta } _ { t } \u0026 \\sim \\frac { \\epsilon _ { t } } { 2 } \\bigl[ G \\left( \\boldsymbol { \\theta } _ { t } \\right) \\bigl( \\nabla _ { \\boldsymbol { \\theta } } \\log p \\left( \\boldsymbol { \\theta } _ { t } \\right) \n",
-        " + \\frac { N } { M } \\sum _ { k = 1 } ^ { M } \\nabla _ \\boldsymbol { \\theta } \\log \\text{GMM}(x_{t_k})\\bigr) + \\sum_\\boldsymbol{\\theta}\\nabla_\\theta G \\left( \\boldsymbol { \\theta } _ { t } \\right) \\bigr]\\\\\n",
+        "\\Delta \\boldsymbol { \\theta } _ { t } \u0026 \\sim \\frac { \\epsilon _ { t } } { 2 } \\bigl[ G \\left( \\boldsymbol { \\theta } _ { t } \\right) \\bigl( \\nabla _ { \\boldsymbol { \\theta } } \\log p \\left( \\boldsymbol { \\theta } _ { t } \\right) +\n",
+        " \\frac { N } { M } \\sum _ { k = 1 } ^ { M } \\nabla _ \\boldsymbol { \\theta } \\log \\text{GMM}(x_{t_k})\\bigr) + \\sum_\\boldsymbol{\\theta}\\nabla_\\theta G \\left( \\boldsymbol { \\theta } _ { t } \\right) \\bigr]\\\\\n",
         "\u0026+ G ^ { \\frac { 1 } { 2 } } \\left( \\boldsymbol { \\theta } _ { t } \\right) \\text { Normal } \\left( \\text{loc}=\\boldsymbol{0} ,\\, \\text{scale}=\\epsilon _ { t }\\boldsymbol{1} \\right)\\\\\n",
         "\\end{align*}$$\n",
         "\n",