feat: load and replace parameters from YAML

**BEFORE**:
```python
from pydantic import BaseModel

class MyExperiment(BaseModel):
'''Training a neural network to do a thing.'''

# identifier after @ indicates a YAML file
network: Network@spk-vih[
# modifications to the YAML contents
shape[1] = 500
layers[1].populations.e.kind = adaptive-LIF
]
loss: Loss@mse
optimizer: Optimizer@adam
dataset: Dataset@default
```

**AFTER**:
```python
from pydantic import BaseModel
from typing import ClassVar, Dict, Any
import yaml
from pathlib import Path

class MyExperiment(BaseModel):
'''Training a neural network to do a thing.'''

# Static storage of YAML modifications
_yaml_mods: ClassVar[Dict[str, Dict[str, Any]]] = {
'network': {
'yaml_id': 'spk-vih',
'modifications': [
('shape[1]', 500),
('layers[1].populations.e.kind', 'adaptive-LIF')
]
},
'loss': {
'yaml_id': 'mse',
'modifications': []
},
'optimizer': {
'yaml_id': 'adam',
'modifications': []
},
'dataset': {
'yaml_id': 'default',
'modifications': []
}
}

# The actual instance fields keep their original types
network: Network # Original type preserved
loss: Loss
optimizer: Optimizer
dataset: Dataset

@classmethod
def load(cls, yaml_dir: str = '.') -> 'MyExperiment':
"""Load and construct instance with YAML data."""
yaml_data = {}

for field, yaml_info in cls._yaml_mods.items():
# Load base YAML
yaml_path = Path(yaml_dir) / f"{yaml_info['yaml_id']}.yaml"
with open(yaml_path) as f:
field_data = yaml.safe_load(f)

# Apply modifications
for path, value in yaml_info['modifications']:
# Split path into parts (handle both dot notation and array indices)
parts = []
current = ''
in_bracket = False
for char in path:
if char == '[':
if current:
parts.append(current)
current = ''
in_bracket = True
elif char == ']':
if current:
parts.append(int(current))
current = ''
in_bracket = False
elif char == '.' and not in_bracket:
if current:
parts.append(current)
current = ''
else:
current += char
if current:
parts.append(current)

# Navigate to target and set value
target = field_data
for part in parts[:-1]:
if isinstance(part, int):
while len(target) <= part:
target.append({})
target = target[part]
else:
if part not in target:
target[part] = {}
target = target[part]

last = parts[-1]
if isinstance(last, int):
while len(target) <= last:
target.append(None)
target[last] = value

yaml_data[field] = field_data

return cls(**yaml_data)
```

The issue names no repository files, tests, or entry points. Start by tracing the proposed MyExperiment.load flow and the @ YAML identifiers shown in the example. Done means the requested YAML-backed fields and path modifications are implemented and verified by tests.

Written by the indexing model from the issue text.