examples: enhance incremental_learning/helmet_detection

[llhuii]

tracking the issues of the incremental learning example.

• [x] the namespace sedna-test should be renamed, default is recommended.
• [x] there exists an useless file named train_data.txt.bak1 in dataset.tar.gz
• [x] there exists an useless directory named base_model/model/ in model.tar.gz
• [ ] need to test hard-example-mining algorithm because of high hard-example storage
• [ ] with bad model eval-worker should report low precision/recall instead of raise exception
• [ ] alleviate the memory usage of training worker

[llhuii]

• [x] in step 2: download base model, actually there are two models: initial-model for training and deploy-model for inference.

[llhuii]

• [x] we should doc which fields should be updated, before just starting the incremental-job, such as the trigger/nodename
• [x] should download the dataset after starting the incremental job.

[llhuii]

• [x] update workerSpec to podTemplate in the incremental job crd.
• [x] fix lib dataset relative-path issue

[TymonXie]

Here the eval worker raised exception with a model trained without initial-model.

[llhuii]

• [ ] In above case eval-worker should report low precision/recall instead of raise exception.

[llhuii]

• [x] remove the Prepare Worker Image section

[llhuii]

train worker consumed lot of memory(about 4GB)

[TymonXie]

inference worker continuous saved examples, result to disk diffusion

[llhuii]

inference worker continuous saved examples, result to disk diffusion

• [ ] need to test hard-example-mining algorithm

[llhuii]

train worker consumed lot of memory(about 4GB)

• [ ] alleviate this memory usage