Add a mean-log-likelihood method to improve the ACT estimation

Suggestion from @colm.talbot. I've then added a number of other tweaks in addition.

bilby/core/sampler/ptemcee.py

@@ -905,21 +905,24 @@ def print_progress(
         convergence_inputs.niterations_per_check * iteration * sampler.nwalkers * sampler.ntemps)
     eval_timing = "{:1.2f}ms/ev".format(1e3 * ave_time_per_check / evals_per_check)
 
-    print(
-        "{}|{}|nc:{}|a0:{}|swp:{}|n:{}<{}|t{}|q:{}|{}".format(
-            iteration,
-            str(sampling_time).split(".")[0],
-            ncalls,
-            acceptance_str,
-            tswap_acceptance_str,
-            nsamples_effective,
-            convergence_inputs.nsamples,
-            tau_str,
-            Q_str,
-            eval_timing,
-        ),
-        flush=True,
-    )
+    try:
+        print(
+            "{}|{}|nc:{}|a0:{}|swp:{}|n:{}<{}|t{}|q:{}|{}".format(
+                iteration,
+                str(sampling_time).split(".")[0],
+                ncalls,
+                acceptance_str,
+                tswap_acceptance_str,
+                nsamples_effective,
+                convergence_inputs.nsamples,
+                tau_str,
+                Q_str,
+                eval_timing,
+            ),
+            flush=True,
+        )
+    except OSError as e:
+        logger.debug("Failed to print iteration due to :{}".format(e))
 
 
 def calculate_tau_array(samples, search_parameter_keys, ci):