[BUG]: usage of camelcase for loss functions in `ml/incr/*`

### Description

In continuation to this conversation [#13333](https://github.com/stdlib-js/stdlib/pull/13333#discussion_r3534668720), we noticed that the packages `ml/incr/binary-classification` and `ml/incr/sgd-regression` follow `camelCase` for mentioning loss functions, where as the current convention is to use `kebab-case`.

Expected outcome is to refactor all the loss functions from using camelcase to kebabcase.

### Related Issues

_No response_

### Questions

No.

### Demo

_No response_

### Reproduction

_No response_

### Expected Results

`ml/incr/binary-classification`
- `modifiedHuber` -> `modified-huber`
- `squaredHinge` -> `squared-hinge`

`ml/incr/sgd-regression`
- `epsilonInsensitive` -> `epsilon-insensitive`
- `squaredError` -> `squared-error`

### Actual Results

```shell

```

### Version

_No response_

### Environments

Node.js

### Browser Version

_No response_

### Node.js / npm Version

_No response_

### Platform

_No response_

### Checklist

- [x] Read and understood the [Code of Conduct](https://github.com/stdlib-js/stdlib/blob/develop/CODE_OF_CONDUCT.md).
- [x] Searched for existing issues and pull requests.

Start by inspecting the ml/incr/binary-classification and ml/incr/sgd-regression package entry points and searching their references for the four camelCase loss-function names. Rename each to its kebab-case form and verify that all package usage and tests consistently use modified-huber, squared-hinge, epsilon-insensitive, and squared-error.

Written by the indexing model from the issue text.