[feat] train: add keep_checkpoint_max config knob

Add TrainConfig.keep_checkpoint_max (default 0 = keep all) so long runs can
cap the number of retained model.ckpt-<step> directories. Documented in
usage/train.md.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: tiankongdeguiji

docs/source/usage/train.md

@@ -63,6 +63,7 @@ LR策略可以支持按epoch更新或者按step更新
 - num_epochs: 训练的epoch数
 - save_checkpoints_steps: 保存模型的步数间隔，保存模型后会做一次评估
 - save_checkpoints_epochs: 保存模型的Epoch数间隔，保存模型后会做一次评估，与save_checkpoints_steps不能同时设置
+- keep_checkpoint_max: 最多保留的最近checkpoint数量，超出后会异步删除最旧的checkpoint，默认0表示全部保留；当exporter_type为best时，会额外保留指标最优的checkpoint
 - fine_tune_checkpoint: 增量训练的checkpoint路径，也可以设置checkpoint目录，将会使用目录下最新的checkpoint
 - fine_tune_ckpt_var_map: 需要restore的参数列表文件路径，文件的每一行是{variable_name in current model}\\t{variable name in old model ckpt}
   - 需要设置fine_tune_ckpt_var_map的情形:

tzrec/protos/train.proto

@@ -70,5 +70,7 @@ message TrainConfig {
     optional uint32 gradient_accumulation_steps = 18;
     // dense gradient clipping config
     optional GradClipping grad_clipping = 19;
+    // maximum number of recent checkpoints to keep; 0 keeps all.
+    optional uint32 keep_checkpoint_max = 20 [default = 0];
     // TBD: qcomm config
 }