Skip consecutive identical history records in UI

While the initial history itself doesn't contain consecutive identical
entries, once you filter for a certain condition it's easy to end up
with spans of identical entries, which isn't at all helpful. Let's just
keep the most recent one.

Author: tecosaur

stdlib/REPL/src/History/resumablefiltering.jl

@@ -259,15 +259,25 @@ Incrementally filter `candidates[1:idx]` in reverse order.
 
 Pushes matches onto `out` until either `maxtime` is exceeded or `maxresults`
 collected, then returns the new resume index.
+
+When there are consecutive entries with identical content and state, all but the
+most recent are skipped.
 """
 function filterchunkrev!(out::Vector{HistEntry}, candidates::DenseVector{HistEntry},
                          spec::FilterSpec, idx::Int = length(candidates);
                          maxtime::Float64 = Inf, maxresults::Int = length(candidates))
+    lightentry(e) = (e.mode, e.content) # for similarity checking
     batchsize = clamp(length(candidates) ÷ 512, 10, 1000)
+    prevaccepted = if isempty(out)
+        HistEntry(:none, DateTime(0), "", 0)
+    else
+        first(out)
+    end |> lightentry
     for batch in Iterators.partition(idx:-1:1, batchsize)
         time() > maxtime && break
         for outer idx in batch
             entry = candidates[idx]
+            lightentry(entry) == prevaccepted && continue
             if !isempty(spec.modes)
                 entry.mode ∈ spec.modes || continue
             end
@@ -294,6 +304,7 @@ function filterchunkrev!(out::Vector{HistEntry}, candidates::DenseVector{HistEnt
             end
             matchfail && continue
             pushfirst!(out, entry)
+            prevaccepted = lightentry(entry)
             length(out) == maxresults && break
         end
     end