proto, runtime/protoiface, internal/impl: add fast-path Merge

Comparing -tags=protoreflect to fast-path:

name                              old time/op    new time/op    delta
pkg:google.golang.org/protobuf/internal/benchmarks goos:linux goarch:amd64
/Clone/google_message1_proto2-12    1.70µs ± 1%    0.30µs ± 1%  -82.64%  (p=0.001 n=7+7)
/Clone/google_message1_proto3-12    1.01µs ± 1%    0.19µs ± 1%  -80.77%  (p=0.000 n=7+8)
/Clone/google_message2-12            818µs ± 8%     141µs ± 6%  -82.78%  (p=0.000 n=8+8)
pkg:google.golang.org/protobuf/internal/benchmarks/micro goos:linux goarch:amd64
EmptyMessage/Clone-12               51.1ns ± 1%    39.3ns ± 3%  -23.03%  (p=0.000 n=7+8)
RepeatedInt32/Clone-12              24.5µs ± 1%     1.1µs ± 3%  -95.64%  (p=0.000 n=8+8)
Required/Clone-12                    978ns ± 1%     132ns ± 2%  -86.46%  (p=0.000 n=8+8)

name                              old alloc/op   new alloc/op   delta
pkg:google.golang.org/protobuf/internal/benchmarks goos:linux goarch:amd64
/Clone/google_message1_proto2-12    1.08kB ± 0%    0.74kB ± 0%  -31.85%  (p=0.000 n=8+8)
/Clone/google_message1_proto3-12      872B ± 0%      544B ± 0%  -37.61%  (p=0.000 n=8+8)
/Clone/google_message2-12            602kB ± 0%     411kB ± 0%  -31.65%  (p=0.000 n=8+8)
pkg:google.golang.org/protobuf/internal/benchmarks/micro goos:linux goarch:amd64
EmptyMessage/Clone-12                96.0B ± 0%     64.0B ± 0%  -33.33%  (p=0.000 n=8+8)
RepeatedInt32/Clone-12              25.4kB ± 0%     3.2kB ± 0%  -87.33%  (p=0.000 n=8+8)
Required/Clone-12                     416B ± 0%      256B ± 0%  -38.46%  (p=0.000 n=8+8)

name                              old allocs/op  new allocs/op  delta
pkg:google.golang.org/protobuf/internal/benchmarks goos:linux goarch:amd64
/Clone/google_message1_proto2-12      52.0 ± 0%      21.0 ± 0%  -59.62%  (p=0.000 n=8+8)
/Clone/google_message1_proto3-12      33.0 ± 0%       3.0 ± 0%  -90.91%  (p=0.000 n=8+8)
/Clone/google_message2-12            22.3k ± 0%      7.5k ± 0%  -66.41%  (p=0.000 n=8+8)
pkg:google.golang.org/protobuf/internal/benchmarks/micro goos:linux goarch:amd64
EmptyMessage/Clone-12                 3.00 ± 0%      2.00 ± 0%  -33.33%  (p=0.000 n=8+8)
RepeatedInt32/Clone-12               1.51k ± 0%     0.00k ± 0%  -99.80%  (p=0.000 n=8+8)
Required/Clone-12                     51.0 ± 0%      18.0 ± 0%  -64.71%  (p=0.000 n=8+8)

Change-Id: Ife9018097c34cb025dc9c4fdd9a61b2f947853c6
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/219147
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
diff --git a/internal/impl/codec_tables.go b/internal/impl/codec_tables.go
index 2c62bda..ef30356 100644
--- a/internal/impl/codec_tables.go
+++ b/internal/impl/codec_tables.go
@@ -20,6 +20,7 @@
 	marshal   func(b []byte, p pointer, f *coderFieldInfo, opts marshalOptions) ([]byte, error)
 	unmarshal func(b []byte, p pointer, wtyp wire.Type, f *coderFieldInfo, opts unmarshalOptions) (unmarshalOutput, error)
 	isInit    func(p pointer, f *coderFieldInfo) error
+	merge     func(dst, src pointer, f *coderFieldInfo, opts mergeOptions)
 }
 
 // valueCoderFuncs is a set of protoreflect.Value encoding functions.
@@ -28,6 +29,7 @@
 	marshal   func(b []byte, v pref.Value, wiretag uint64, opts marshalOptions) ([]byte, error)
 	unmarshal func(b []byte, v pref.Value, num wire.Number, wtyp wire.Type, opts unmarshalOptions) (pref.Value, unmarshalOutput, error)
 	isInit    func(v pref.Value) error
+	merge     func(dst, src pref.Value, opts mergeOptions) pref.Value
 }
 
 // fieldCoder returns pointer functions for a field, used for operating on