perf: add uri.escape/unescape test

Added a simple benchmark for URI escape/unescape.

Part of tarantool#3682

NO_DOC=documentation is not required for performance test
NO_CHANGELOG=performance test
NO_TEST=performance test

perf/lua/uri_escape_unescape.lua

@@ -0,0 +1,43 @@
+-- It is recommended to run benchmark using taskset for stable results.
+-- taskset -c 1 tarantool uri_escape_unescape.lua
+
+local uri = require("uri")
+local clock = require("clock")
+local t = require("tarantool")
+
+local _, _, build_type = string.match(t.build.target, "^(.+)-(.+)-(.+)$")
+if build_type == "Debug" then
+    print("WARNING: tarantool has built with enabled debug mode")
+end
+
+local escape_opts = {
+    unreserved = uri.unreserved("A-Z"),
+    plus = true,
+}
+-- Alternate reserved and unreserved symbols plus a space symbol.
+local str = "A$B$C$D$E$F$G$H$I$J$K$L$M$N$O$P$Q$R$S$T$U$V$W$X$Y$Z$ #"
+local encode_sample = string.rep(str, 10^3)
+local decode_sample = uri.escape(encode_sample, escape_opts)
+
+-- JIT compilation starts work after detecting "hot" paths in loops and
+-- function's calls. It potentially will make execution time of first
+-- iterations slower that next. To avoid such effect we can introduce
+-- "warmup" iterations, disable JIT at all or make JIT compilation
+-- aggressive from the beginning.
+jit.opt.start("hotloop=1", "hotexit=1", "recunroll=1", "callunroll=1")
+
+local cycles = 10^3
+local start = clock.monotonic()
+for _ = 1, cycles do
+    uri.escape(encode_sample)
+end
+local encode_time = cycles / (clock.monotonic() - start)
+
+start = clock.monotonic()
+for _ = 1, cycles do
+    uri.unescape(decode_sample)
+end
+local decode_time = cycles / (clock.monotonic() - start)
+
+print(("uri.escape   %.2f  runs/sec"):format(encode_time))
+print(("uri.unescape %.2f  runs/sec"):format(decode_time))