[BUG] Unexpected behavior when CheckpointSaver's parameter max_history is 1

[chenshen03]

Describe the bug
When the parameter max_history in the CheckpointSaver.py is set to 1, the checkpoint is saved for each epoch.

To Reproduce
Steps to reproduce the behavior:

import torch
import timm
from timm.utils import CheckpointSaver

model = timm.models.xception()
optimizer = torch.optim.Adam(model.parameters())

saver = CheckpointSaver(model, optimizer, max_history=1)
saver.save_checkpoint(0, metric=0.5)
saver.save_checkpoint(1, metric=0.5)
saver.save_checkpoint(2, metric=0.6)
saver.save_checkpoint(3, metric=0.7)
saver.save_checkpoint(4, metric=0.8)
saver.save_checkpoint(5, metric=0.7)

Expected behavior
The checkpoint of each epoch will be saved.

Screenshots

Desktop (please complete the following information):

• OS: Ubuntu 18.04
• This repository version [e.g. pip 0.3.1 or commit ref]
• PyTorch version w/ CUDA/cuDNN [e.g. from conda list, 1.7.0 py3.8_cuda11.0.221_cudnn8.0.3_0]

Additional context
Add any other context about the problem here.

[rwightman]

@chenshen03 thanks, I'll look at this soon, it is supposed to work but it's not something i've tested.