feat: data collection for link click statistics

[kylerwsm]

Problem

This PR aims to start-off data collection for our upcoming link click statistics feature.

Closes #168.

Solution

We do this by collecting and storing data directly to dedicated tables, each representing a statistic we will show to users. Additionally, in this implementation, badly formatted data will not be written to the statistics databases.

More information on how the each statistic is implemented:

Daily Statistics

• Uses a composite key, a combination of the short url and date
• Date is in Singapore time
• Index is unique and uses the shortUrl and date fields

Weekday Statistics

• Uses the short url as the primary key.
• Each day is represented by an integer between 0 (Sunday) to 6 (Saturday). This representation aligns with JavaScript Date objects.
• Index is unique and uses the shortUrl, weekday, and hours fields

Device Statistics

• Uses the short url as the primary key.
• Makes use of the ua-parser-js library to parse the user agent string to a device type.
• Limitation: Desktop devices do not identify their device types in user agent string, unlike mobile and tablet devices. I have then defaulted user agent strings that do not identify their device types to be desktop devices.
• Index is unique and uses only the shortUrl field

[yong-jie]

have confirmed with @kylerwsm that data collection is presently working when tested on staging

[liangyuanruo]

as commented. it would be good if we can try creating the tables without indexes, then run an EXPLAIN SELECT ... and make sure that we're still able to rely on the composite primary keys without doing any table scans here. indexes result in additional write overhead, best if the composite PKs can provide indexed lookup.