[quant] Add back README.md for backend_config (#70964)

Summary:
Pull Request resolved: #70964

Accidently deleted before, adding this back. We'll make this more complete after
the structure is finalized

Test Plan:
no test needed

Imported from OSS

Reviewed By: dagitses

Differential Revision: D33470738

fbshipit-source-id: 00459a4b00514d3d0346de68788fab4cad8a5d12

Author: jerryzh168

torch/ao/quantization/fx/backend_config/README.md

@@ -0,0 +1,20 @@
+The patterns are we matching against is float modules types, functional operators and pytorch operators in reverse order:
+```
+operator = module_type | functional | torch op | native op | MatchAllNode
+Pattern = (operator, Pattern, Pattern, ...) | operator
+```
+where the first item for Pattern is the operator we want to match, and the rest are the patterns for the arguments of the operator.
+For example, pattern (nn.ReLU, (operator.add, MatchAllNode, (nn.BatchNorm2d, nn.Conv2d))) would match the following graph:
+```
+tensor_1            tensor_2
+ |                    |
+ *(MatchAllNode)  nn.Conv2d
+ |                    |
+ |             nn.BatchNorm2d
+ \                  /
+  -- operator.add --
+         |
+      nn.ReLU
+```
+
+we’ll match the last node as the anchor point of the match, and we can retrieve the whole graph by tracing back from the node, e.g. in the example above, we matched nn.ReLU node, then node.args[0] is the operator.add node.