[Ansor] Support multiple output ops and fix Python API printing

[comaniac]

As I'm trying to use Ansor to tune the operators generated from te.gradient, I fixed some issues in this PR so that now Ansor can tune the backward ops (regardless the performance). Detail change list:

• [Multiple Output] Create a schedule with all outputs instead of the first output.
• [Python API Printing] Change tvm.thread_axis to te.thread_axis.
• [Python API Printing] Add prefix to all iterators to avoid name conflict; otherwise we may refer to the wrong iterator with the same name hint (e.g., ax0). For example:

ax0, ... = tuple(pad_temp_data_grad.op.axis) + ...
...
ax0, ... = tuple(pad_temp_shared.op.axis) + ...
s[pad_temp_data_grad].split(ax0, factor=4)

where ax0 in split should refer to the ax0 from pad_temp_data_grad, but it has been overridden by pad_temp_shared after cache_read. This PR improves CleanName by providing an optional prefix so that we can differentiate those iterators by their stages.

cc @merrymercy @jcf94

[comaniac]

Current issue: te.InferBound failed for schedules with multiple outputs. Investigating.

[comaniac]

Two more bug fixing:

1. Only specify output tensors when creating a new schedule; otherwise InferBound may crash.

2. Update compute DAG attributes such as access analyzer and ops after layout rewriting (cc @minminsun ).

All auto_scheduler tests are now passed.

[merrymercy]

Why did it work before without the fix?

[comaniac]

I wish I know the answer... For the ApplySteps in compute DAG, previously we simply put all non-placeholder ops to the list and take the last op (assuming it is the only output) to create a new schedule. After using access_analyzer.is_output to check what ops should be put to the list, is_output throws exceptions due to the out-of-date is_output map. That's why I found this issue. Maybe most use cases so far don't enable layout rewrite?

[tqchen]

Thanks @comaniac @merrymercy