support explicit model

[dirty49374]

In this patch, format() function returns the model used in render() call not the rendered string.
( I did it because I want to write a wrapper javascript generator and I wanted to reuse current one. )

changes are like this.

// before
export namespace EnumFormatter {
    export function format(enumDescriptor: EnumDescriptorProto, indentLevel: number): string {
        let enumName = enumDescriptor.getName();
        let values: { [key: string]: number } = {};
        enumDescriptor.getValueList().forEach(value => {
            values[value.getName().toUpperCase()] = value.getNumber();
        });
        return TplEngine.render('partial/enum', {
            indent: Utility.generateIndent(indentLevel),
            enumName: enumName,
            values: values,
        });
    }
}
// after
export namespace EnumFormatter {
    export interface EnumModel {
        indent: string;
        enumName: string;
        values: { [key: string]: number };
    }
    export function format(enumDescriptor: EnumDescriptorProto, indent: string): EnumModel {
        let enumName = enumDescriptor.getName();
        let values: { [key: string]: number } = {};
        enumDescriptor.getValueList().forEach(value => {
            values[value.getName().toUpperCase()] = value.getNumber();
        });
        return {
            indent,
            enumName: enumName,
            values: values,
        };
    }
}

// before
{{#each message.formattedEnumListStr}}
{{{this}}}
{{/each}}

// after
{{#each message.formattedEnumListStr}}
{{{render 'partial/enum' this}}}
{{/each}}

this commit also contains hbs template cache.

[agreatfool]

Just merged & latest 2.5.4 has been released. Thank you guys for the great enhancement.