move dynamic linear to main directory (#180)

Summary:
perf and accuracy are great, so makes more sense to have this at the same level of importance as `Float8Linear`

Pull Request resolved: #180

Test Plan:
```
./test/test_everything.sh
```

Reviewed By: drisspg

Differential Revision: D52648574

Pulled By: vkuzo

fbshipit-source-id: cc304c56360731cc8fe4b8bae5418683be75c684

Author: vkuzo

README.md

@@ -143,7 +143,7 @@ We are working on a new FSDP implementation that uses per-parameter sharding tha
 # code tips
 
 * `float8_experimental/float8_linear.py` - `Float8Linear` (main user facing entry point for delayed scaling)
-* `float8_experimental/dynamic_linear/dynamic_linear.py` - `Float8DynamicLinear` (main user facing entry point for dynamic scaling)
+* `float8_experimental/float8_dynamic_linear.py` - `Float8DynamicLinear` (main user facing entry point for dynamic scaling)
 * `float8_experimental/float8_tensor.py` - `Float8Tensor`, which allows `Float8Linear` to abide by the `x.dtype == x.grad.dtype` restriction
 * `float8_experimental/tp_linear.py` - `Float8ColumnParallelLinear` / `Float8RowParallelLinear` (TP/SP versions of float8 linear)
 

float8_experimental/dynamic_linear/__init__.py

@@ -8,7 +8,7 @@
 
 import torch
 import torch.distributed as dist
-from float8_experimental.dynamic_linear import Float8DynamicLinear
+from float8_experimental.float8_dynamic_linear import Float8DynamicLinear
 from float8_experimental.float8_linear import Float8Linear
 
 from float8_experimental.float8_utils import amax_history_to_scale, tensor_to_amax