load_partition() fails for heterographs whose edge relations are not unique

## 🐛 Bug

load_partition() fails to partition a heterogeneous graph whose edges have the same relation but are different.

## To Reproduce

Steps to reproduce the behavior:

1. Create a heterograph (**Creating a Heterogeneous Graph** section): https://docs.dgl.ai/en/0.6.x/guide/graph-heterogeneous.html
```python
import dgl
import torch as th
# Create a heterograph with 3 node types and 3 edges types.
graph_data = {
('drug', 'interacts', 'drug'): (th.tensor([0, 1]), th.tensor([1, 2])),
('drug', 'interacts', 'gene'): (th.tensor([0, 1]), th.tensor([2, 3])),
('drug', 'treats', 'disease'): (th.tensor([1]), th.tensor([2]))
}
g = dgl.heterograph(graph_data)
```
2. Run ``` dgl.distributed.partition_graph(g, "example", out_path="output/", num_parts=2) ``` and get the following error:
```python
>>> dgl.distributed.partition_graph(g, "example", out_path="output/", num_parts=2)
Convert a graph into a bidirected graph: 0.000 seconds
Construct multi-constraint weights: 0.000 seconds
[18:35:34] /home/tiger/usr_name/dgl/src/graph/transform/metis_partition_hetero.cc:78: Partition a graph with 10 nodes and 10 edges into 2 parts and get 1 edge cuts
Metis partitioning: 0.000 seconds
Reshuffle nodes and edges: 0.001 seconds
Split the graph: 0.000 seconds
Construct subgraphs: 0.004 seconds
Traceback (most recent call last):
File "", line 1, in
File "/home/tiger/anaconda3/envs/usr_name/lib/python3.8/site-packages/dgl-0.8-py3.8-linux-x86_64.egg/dgl/distributed/partition.py", line 631, in partition_graph
inner_etype_mask = inner_etype == g.get_etype_id(etype)
File "/home/tiger/anaconda3/envs/usr_name/lib/python3.8/site-packages/dgl-0.8-py3.8-linux-x86_64.egg/dgl/heterograph.py", line 1261, in get_etype_id
etid = self._etypes_invmap.get(self.to_canonical_etype(etype), None)
File "/home/tiger/anaconda3/envs/usr_name/lib/python3.8/site-packages/dgl-0.8-py3.8-linux-x86_64.egg/dgl/heterograph.py", line 1153, in to_canonical_etype
raise DGLError('Edge type "%s" is ambiguous. Please use canonical edge type '
dgl._ffi.base.DGLError: Edge type "interacts" is ambiguous. Please use canonical edge type in the form of (srctype, etype, dsttype)
```

## Expected behavior

Partition should work for graphs whose canonical edges are distinct, and it shouldn't depend on edge relation.

## Environment

- DGL Version (e.g., 1.0): **0.8**
- Backend Library & Version (e.g., PyTorch 0.4.1, MXNet/Gluon 1.3): **PyTorch 1.9.0**
- OS (e.g., Linux): **Debian**
- How you installed DGL (`conda`, `pip`, source): **source**
- Build command you used (if compiling from source): **successful**
- Python version: **3.8.11**
- CUDA/cuDNN version (if applicable): **NA**
- GPU models and configuration (e.g. V100): **NA**
- Any other relevant information:
1. The simplest solution is to use different edge relation names. However, this may not be a good solution, because this makes the concept of canonical edge types the same as edge types.
2.
The problem seems to happen whenever (in partition.py)
```python
for etype in g.etypes:
etype_id = g.get_etype_id(etype)
```
It is related to [to_canonical_etype()](https://github.com/dmlc/dgl/blob/738b75f41e5d3229e5ccda52d76e1297d7b0520d/python/dgl/heterograph.py#L1094). One can change ```g.etypes``` to ```g.canonical_etypes```. But this has other repercussions when saving meta graph data into json and reloading the graph. Essentially the edge relation has to be distinct.